5th World Symposium of Computers in Obstetrics and Gynecology
(ISCOG '97)

Return to the Table of Contents

 

ANALYSIS OF FETAL HEART RATE

R Dawkins, T Dawkins, TKH Chung, DS Sahota, AMZ Chang

Department of Obstetrics & Gynaecology, The Chinese University of Hong Kong, Hong Kong

To find a single word to define the whole field of Fetal Heart Rate analysis one is inevitably drawn to "inconsistency". There is inconsistency in the interpretation of fetal heart rate traces both between individuals and in the same individual at different times. There is inconsistency in the application of outcome measures of intrapartum hypoxia. Apgar score, cord blood gas analysis, neonatal encephalopathy, spasticity and mental retardation have all been used in varying combinations and definitions. Finally there is inconsistency in the correlation between these outcome measures and clearly defined Fetal Heart Rate abnormalities. Against the background of evidence based medicine Fetal Heart Rate analysis is at risk of being discredited, potentially solely because of the inconsistencies in interpretation and outcome measures. Consistency can only be achieved by precise definitions and strict adherence to those definitions. Reasonably precise definitions for Fetal Heart Rate analysis are in place. Addressing inconsistent interpretation and outcome measures remain a priority. For several decades efforts have been made to computerise the analysis of Fetal Heart Rate. The advantages of consistent reproducible interpretation have driven this research. Antenatal systems such as the System 8000 have achieved considerable success, but as yet real time intrapartum analysis has proved more elusive. The major problem is recognition of the baseline. This stems from two sources. Firstly the definition. FIGO have defined baseline fetal heart rate as the mean level of the fetal heart rate when this is stable, accelerations and decelerations being absent, determined over a period of 5 to 10 minutes and expressed in beats/min. This is a circular argument, ill-advised at the best of times, but a particular problem when trying to devise a computer algorithm to recognise the baseline in real time. Secondly, intrapartum Fetal Heart Rate analysis is commonly complicated by frequent deviations from the baseline and prolonged episodes of missing data. In a fresh attempt to assess the value of intrapartum Fetal Heart Rate analysis a group of Obstetricians with an interest in programming are developing a computer program for real time Fetal Heart Rate analysis. Several different approaches for baseline recognition were tried prior to agreement on the following. The baseline algorithm utilises a gate to exclude accelerations and decelerations, accumulates the most recent 6 minutes of stable fetal heart rate in a "baseline array", determines the modal range of the "baseline array" using a class width of 10 beats per minute and returns the mean of the modal range to plot the baseline. A reset mechanism allows the baseline to rapidly reset should the fetal heart rate stabilise outside the gate. The presentation will give further detail of the baseline algorithm and demonstrate the program in action.

Return to the Table of Contents

  

COMPUTERISED ESTIMATION OF THE BASELINE FETAL HEART RATE IN LABOUR: THE LOW-FREQUENCY LINE

TKH Chung, R Dawkins, DS Sahota, MS Rogers, AMZ Chang

Department of Obstetrics & Gynaecology, The Chinese University of Hong Kong, Hong Kong

The objective was to assess computerized algorithm for the estimation of the fetal heart rate baseline (low-frequency line) during labour. A retrospective observational study was performed. Fetal heart rate signals were obtained from women in labour using the Nottingham fetal ECG monitor. The computerized algorithm for the baseline estimation was developed for intra-partum applications has been previously described. Evaluation was carried out on sixty cardiotocographic recordings by 12 experts and by the computer. These estimates were compared with those obtained from the computerized system using paired differences and intraclass correlation. The study showed that it is possible to produce a low-frequency line from data obtained from intrapartum records. The system could not estimate the low-frequency line in 4 records, whereas experts were also unable to estimate between 1 to 7 tracings. The 95% C.I. for the paired differences between computer and experts was -12 to 15 bpm, whereas between the experts this was -10 to 10. With the exception of one expert, there was a high concordance between experts and between computer and experts (intraclass correlation >0.9). The performance of this computerized algorithm cannot be distinguished from that of experienced clinicians. There were no significant differences between baseline values obtained by the computerized algorithm and those by the clinicians.

Return to the Table of Contents

 

COMPUTERISED CTG ANALYSIS –RELATIONSHIP TO FETAL SCALP PH

*BK Strachan, *WJ Van Wijngaarden, *DK James, **DS Sahota, **AZM Chang

*Department of Obstetrics and Gynaecology, University of Nottingham, UK **Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, HK

Part of the validation of a computerised CTG analysis system is to compare its performance with experts. A further step is to compare the analysis of the CTG components as measured by the analysis program and compare these to outcome measures. We present the relationship of these components to fetal scalp pH. From a large database of 800 labours recorded as part of the Nottingham fetal ECG project a database of 181 fetal blood samples taken from 140 women in labour were identified. The 30 minutes of CTG prior to the FBS was analysed. Acidaemia was defined as a scalp pH of less than 7.2. For each CTG component a receiver operator curve was calculated to identify those components which had a significant predictive value for a low pH. Tachycardia, presence of accelerations proved not to be predictive for a low pH. Bradycardia was rarely present prior to fetal scalp sampling and therefore could not be evaluated. Low variability had a very weak association. Decelerations as measured by total deceleration area, particularly the area of deceleration occurring after a contraction had the best relationship to a low pH.

Return to the Table of Contents

  

COMPUTERISED TRANSFORMATION OF CTG PAPER TRACING INTO ITS DIGITAL EQUIVALENTS

DS Sahota, PM Yuen, TKH Chung and AMZ Chang

Department of Obstetrics & Gynaecology, Prince of Wales Hospital, The Chinese University of Hong Kong, Hong Kong

Visual interpretation of the cardiotocogram (CTG) is known to be difficult, uncertain and inconsistent and has lead to the development of computerised CTG analysis. However, most of the fetal heart rate tracings are recorded in paper form and these must first be converted into digital values if they are to be retrospectively analysed by computers. We describe a method by which CTG tracings can be converted into digital form suitable for such forms of analysis. A scan image of each major manufacturers’ CTG paper was used to identify the colour clusters associated with the background grid and static text. All images were scanned using a horizontal resolution of 152 dpi so that a single pixel represented a one second epoch. The images were then rotated so that the grids were aligned in the same horizontal planes. Static grid markings were removed using a logistic contrast enhancement transformation. Hand written annotations on the trace were removed by manual editing technique. This leaves only the fetal heart rate and tocographic tracings. The tracings were converted into digital values by first determining the pixel range for each epoch. The minimum and maximum pixel values were then digitally filtered and re-scaled to digital values according to the original heart rate scale on the paper. Forty 18-minute segments of intrapartum CTG tracing were recorded on paper and also as digital values in one second epochs directly from the fetal monitor. The method was validated by comparing the difference between the fetal heart rate values obtained directly from the fetal monitor with those obtained indirectly from the paper recordings The maximum mean difference between the actual and derived heart rate values was less than 1.17 beats per minute. For all segments, the actual values were greater than the derived values with an overall mean difference of less than a beat per minute (0.76). The methodology described allows a quick, easy and accurate means of converting CTG tracings into their digital equivalent and has been adapted for use with photocopied CTG paper tracing. This permits the current large and extensive CTG database of recordings to be analysed by computer based CTG interpreters

Return to the Table of Contents

 

IMPROVING INTRAPARTUM SURVEILLANCE: (I) FETAL ECG MORPHOLOGY AND TIME INTERVALS

WJ van Wijngaarden, BK Strachan and DK James

Department of Obstetrics & Gynaecology, University of Nottingham, Queens Medical Centre, Nottingham NG7 2UH, UK

Objectives: Both time interval and morphological changes of the FECG have been seen to accompany fetal compromise during labour. During hypoxaemia the normally negative PR time interval - FHR relationship can be predominantly positive as well as completely positive during 20 consecutive minute epochs. The first aim of this pilot study was to establish whether the use of a percentage of the time during a 20 minute period this relationship is positive is superior in the detection of fetal compromise to the use of a completely positive relationship during all of the 20 minute period. Reports on changes in the T/QRS ratio during fetal compromise using a preset level to depict abnormality are equivocal. The second aim of this pilot study was to assess whether an individual cut-off level has a potential in improving the detection of fetal compromise. Study Design: A retrospective analysis of 20 intrapartum FECG recordings of cases recruited in the Nottingham multicentre EFM versus EFM plus FECG intrapartum monitoring trial. Results: A positive PR interval - FHR relationship for 96.7% of a 20 minute epoch detects fetal compromise better than when positive for 100% of such a period without an obvious increase in adverse effects. An increase in the T/QRS ratio over the 99.5th centile for 2 consecutive minutes calculated on an individual basis, discriminates best between biochemically compromised and non-compromised fetuses (cut-off compromise/non-compromise: pH = 7.15, Base Excess = -10mM). The addition of these FECG waveform analyses to EFM monitoring has the potential to improve the detection of compromised fetuses. Conclusion: The detection of fetal compromise during labour could be improved by using a percentage of time (96.7%) rather than all of the time the PR - FHR relationship is positive over a 20 minute epoch. T/QRS ratio changes with individually calculated criteria for abnormality may have some benefit in the detection of fetal compromise but the best format remains to be established. Larger studies are required to evaluate these findings.

Return to the Table of Contents 

OBJECTIVE DECISION MAKING WITH OUTCOME PROBABILITIES OBTAINED BY A NEURAL NETWORK COMPUTER SYSTEM

K Maeda1, M Utsu1, Y Noguchi2, T Hamada2, K Mariko2, F Matsumoto2

1Department of Obstetrics & Gynecology, Seirei Hamamtsu & Mikatahara Hospital; 2Department Applied Physics, National Defense Academy, Japan

An experts’ knowledge system developed in 1980 has been used in fetal monitoring for 15 years. Its new application in a TOITU telemetry central monitor showed its reliability and usefulness in general intrapartum fetal surveillance. In the other hand, we have recently studied more objective decision making by the application of artificial neural network analysis. The system is composed of a personal computer and a back propagation software written in C language with 40 input, 30 intermediate and 3 output units. The computer system was trained with the 8 FHR parameters including baseline heart rate, variability amplitude, presence of sinusoidal pattern, number of decelerations, 4 deceleration parameters including its duration, dip heart rate, recovery time and lag time, obtained in 20 typically normal, suspicious and pathological cases. Most FHR data were obtained by our conventional FHR analysis program, and they were input into the computer after their automatic changes into 16 steps in the normalization. The data were stored in the computer memory for the use in the network training which was repeatedly done for 10,000 times. The probabilities to be normal, suspicious or pathological outcome were obtained by the trained computer. Long term results obtained in 50 minutes were correct in 93% of 29 cases comparing them to the combined analysis of CTG, our conventional Program and umbilical pH. Short term probabilities in 15 min were obtained every 5 min continuously, and the results were successively displayed in a trendgram. New 10 cases were evaluated by the trendgram. All results were agreeable when they were compared to the CTG and clinical outcomes. A normal case continuously showed 100% normal outcome probabilities. An abnormal CTG of late deceleration and the loss of variability fully showed 100% pathological probabilities. Suspicious outcome was estimated by the mixed normal and pathological probabilities and by the high probability of suspicious outcome. Neural network computer shows the decision on fetal outcome by the probability, whereas detailed FHR changes and alarm signs are shown by our conventional experts’ program. Therefore, combined use of the two systems forming a new parallel-hybrid system will be more useful in clinical application in future use. Although we have used keyboard input, the procedure will be fully automatic in future systems.

Return to the Table of Contents

GENERALIZED REGRESSION NEURAL NETWORK CLASSIFICATION OF CARDIOTOCOGRAPHIC PATTERNS

BE Rosen, T Bylander

Division of Computer Science, University of Texas at San Antonio, 6900 North Loop 1604 West, San Antonio TX 78249, USA

Cardiotocography has become a primary method of evaluation of fetal wellbeing. We investigated the ability of Generalized Regression Neural Networks (GRNNs) to predict fetal outcome from feature patterns associated with difficult to classify cardiotocograms. Each of the 153 patterns contained 37 attributes consisting of the CTG features and auxiliary information. The patterns also contained a single attribute indicating the fetal outcome, classified as healthy (60.8%), problem (26.8%) or death (12.4%). Using leave-one-out crossvalidation, initial experiments show the GRNN model had a prediction accuracy 81%, and had a 3% major misclassification rate (e.g. a prediction of a "healthy" when the outcome was "death" or a prediction of a "death" when the outcome was "healthy"). In contrast, the C4.5 decision trees method classified only 71% of the patterns correctly, and had a major misclassification rate of 6%. When only high confidence levels are considered (i.e. above 90%), the GRNN model correctly predicts fetal outcome with 92% accuracy and has a 2% major misclassification error. Our results indicate the GRNN approach has the potential for aiding the physician in the diagnosis of fetal wellbeing.

Return to the Table of Contents

VARIABLE TRANSFORMATION IN PREDICTION MODEL DEVELOPMENT: THE ROLE OF GENERALIZED ADDITIVE MODELLING

LY Hin, TK Lau, MS Rogers, AMZ Chang

Department of Obstetric & Gynecology, Prince of Wales Hospital , The Chinese University of Hong Kong, Hong Kong

In clinical medicine, prediction models are used to estimate risk and prognosis. Many of the predictors in these models are continuous measurements which are often dichotomized to simplify the models and to facilitate decision making. The choices of these thresholds were empirical and usually based on published epidemiological studies that may not be representative of the local population due to difference in demographic characteristics and disease pattern. Therefore, these thresholds may be inaccurate, and weakens the prediction models. In addition, generalized linear models (GLM) used for multivariate analyses in those studies require assumption of relationships between the response and each predictor. This prevents estimation of the true response-predictor relationships that are vital in determining the number of thresholds that should be chosen for dichotomization of each predictor. The generalized additive models (GAM) allow non-parametric estimation of the response-predictor relationships with concurrent adjustment for confounding variables. This enables identification of the number of interval(s) along each of the predictors associated with positive logit values, ie., higher risk of developing the measured response, and the identification of threshold(s) that bound these intervals. This technique was applied to predict intrapartum Caesrean deliveries among 11475 term singleton pregnancies in cephalic presentation. The sample was randomly assigned to 2 groups. Two models (LRHUESTON & LRGAM) were developed from one group and validated on the other, and vice versa. The results were pooled and ROC curves1 generated from these two models were compared using DeLong’s2 approach.. Using the empirical thresholds (18 years old and 35 years old) chosen by Hueston3 for maternal age at delivery, we dichotomized this predictor and developed a model (LRHUESTON) for each group. Using GAM, we demonstrated that the relationship between intrapartum Caesarean delivery and maternal age is monotonous, and the threshold for dichotomization was chosen at 28.8 and 28.6 for the two groups respectively, and developed a model (LRGAM) for each group. Area under ROC for LRHUESTON and LRGAM were 0.75 and 0.77 respectively, and LRGAM predicts significantly more accurate than LRHUESTION (z = 5.75, P < 0.001).

Return to the Table of Contents

A NOVEL SYSTEM FOR PROVIDING COMPATIBLE BLOOD TO PATIENTS DURING SURGERY: "SELF-SERVICE" ELECTRONIC BLOOD BANKING BY NURSING STAFF

G Cheng

Department of Medicine, Chinese University of Hong Kong

Background: A good blood bank must be able to provide compatible blood units promptly to operating room patients with minimal wastage. A "self-service" by nursing staff blood banking system that is safe, efficient, and well-accepted has been developed. Study Design and Methods: Specific blood units are no longer assigned to surgical patients who have a negative pretransfusion antibody screen, irrespective of the type of surgery. A computer-generated list of the serial numbers of all group-identical blood units currently in the blood bank inventory is provided for each patient. The units themselves are not labeled with a patient’s name. The group O list will be provided for group O patients, the group A list for group A patients, and so forth. Should the patient require transfusion during surgery, the operating room nurses go to the refrigerator, remove any group-identical unit, and check the serial number of the unit against the serial numbers on the patient’s list. If the serial number is on that list, the blood bank will accept responsibility for compatibility. The system was implemented in 1995. Results: Since implementation, a total of 12,154 patients have undergone operations at this hospital. One hundred and thirty-two patients received more than 10 units of red cells each. There were no transfusion errors. The crossmatch-to-transfusion ratio was reduced from 1.67 to 1.12. Turnaround time for supplying additional or urgent units to patients in operating room was shortened form 33 to 2.5 minutes. There was no incidence of a blood unit’s serial number not being on the list. Work by nurses and technical staff was reduced by nearly 50 percent. Conclusion: The "self-service" (by nursing staff) blood banking system described is safe and efficient. It saves staff time and can be easily set up.

Return to the Table of Contents

THE DEVELOPMENT OF A HYBRID EXPERT SYSTEM FOR THE INTERPRETATION OF FETAL ACID-BASE STATUS

TN Leung, DS Sahota, AMZ Chang

Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, Prince of Wales Hospital, Hong Kong

An expert system was developed using a combination of logistic transformations, back-propagation neural networks (BNN) and decision algorithms. These mathematical techniques were used to reject samples unsuitable for interpretation or to categorise them according to their metabolic status.	 A sample rejection algorithm was designed to reject those samples where individual variables values were clearly out of physiological range or if patterns of different measurement variables such as pH and pO2 and pCO2 were physiologically incoherent. A BNN trained using a set of prescribed patterns was created to classify samples according to their metabolic status. Outputs from this BNN are used to indicate if a sample is either acidotic or alkalotic as well as to indicate atypical patterns which are combinations of variable values not in error but which could not be interpreted. The oxygen tension was logistically transformed and interpreted separately. Output indices were produced by the expert system using only those BNN outputs relating to acidosis. The type of acidosis was determined by summing the metabolic acidosis and respiratory acidosis terms, whilst the difference of these two outputs reflected the nature of the balance between metabolic and respiratory acidosis. The expert system was tested on a database of 2174 samples collected by transferring digital measurement values from blood samples analysed by a Corning 238 blood gas analyser. Information downloaded was stored on a computers hard disk and subsequently interpreted by the expert system. Of the 2174 samples analysed 88 samples would have been rejected as erroneous; 13 because of an out-of-range pH alone ( ³ 7.48); 73 because more than one measurement was marginally out of range, and two because the relationship between measurements was inconsistent. A total of 527 samples would have diagnosed as being acidotic; of these, 139 were respiratory , 114 mixed and 274 metabolic

Return to the Table of Contents

 

EXPERT SYSTEMS - FORWARDS, BACKWARDS OR SIDEWAYS?

T Chard

St Bartholomew’s Hospital, London, UK

Expert systems have been a major topic at all previous meetings of both ISCONG and similar local societies. At all of these it has been confidently predicted that Expert Systems would enter widespread use in Obstetrics and Gynaecology within the succeeding five years. Yet the reality is that Expert Systems are still limited to a few topics, usually in association with well-defined investigational tools such as CTGs or ultrasound. This is not for lack of published systems in almost all areas of Obstetrics and Gynaecology nor is it due to any deficiencies in the available technology, nor is it because Expert Systems have not been demonstrated to perform as well or better than most humans. The probable explanation is that a free-standing Expert System is impractical in the absence of a data collection system operated from the same workstation. Only when such systems have become a routine part of clinical practice in Obstetrics and Gynaecology (as they are, for example, in areas such as General Practice) will Expert Systems have a significant impact on day-to-day clinical activity. At that time other constraints may also become apparent, in particular, the potential legal liabilities which surround what will inevitably be deemed as an investigational tool equivalent to all other clinical technologies.

Return to the Table of Contents  

IMPROVING CERVICAL SCREENING: THE ROLE OF NEW TECHNOLOGIES

SKC Ng

President, New System International Ltd, Hong Kong

Increasing the detection of significant abnormalities on cervical cytologic smears with the use of computers has recently been made possible with the application of neural network technology. The PAPNETâ system searches for and identifies the most suspicious-appearing cells on each cervical smear and displays them on a video monitor for interpretation and diagnosis by a cytologist. The system is intended to increase screening accuracy and decrease the incidence of false-negative Pap smears by targeting those abnormalities that are most likely to go undetected by microscopic screening. The system has been designed to be particularly sensitive in detecting abnormal cells that are few in number (< 100 abnormal cells per slide) or small cells (< 15 microns). The US FDA has approved data which show that cytologists can detect significantly more abnormal smears (at least 7 times) using the PAPNET system as compared to rescreening the same number of smears using a microscope alone. In addition, three independent investigators reported that PAPNET analysis of smears originally interpreted as ASCUS can identify evidence of squamous intraepithelial lesions (SILs) in a significant number of cases. Finally, worldwide data and experience show that this technology may also have applications as a primary screening device for the detection of abnormalities on cervical smears.

Return to the Table of Contents

 

EVALUATION OF NUMERIC MEASUREMENTS - AUTOMATIC QUALITY ASSURANCE IN FETAL HEART RATE INTERPRETATION

Zoltan Takacs

Hewlett-Packard GMbH, Germany

Hewlett-Packard’s OB TraceVue system consists of many separate software modules. One of these is its alerting software module. This software enables clinicians to reliably measure and analyze acceleration, deceleration, and contraction patterns of fetal heart rate traces (Cardiotocogrammes). This session discusses how the alerting software measures these patterns accurately. Features such as amplitude, duration and area, with predetermined values, are entered into a trace generator. The trace generator produces: a numeric master-trace record an artificial-trace The master-trace record passes straight to an external quality assurance software module. The artificial-trace passes first to the OB TraceVue alerting software for evaluation. It is analyzed in exactly the same way as genuine patient trace data. The altering software describes the artificial-trace as a numerical record. Finally, this numeric record passes to the external quality assurance software module. The quality assurance software compares the two records numerically. The accuracy of OB TraceVue’s alerting algorithm is confirmed when the artificial-trace record falls within an acceptable tolerance of the master-trace record.

 

Return to the Table of Contents

SELF ORGANISING MAP NEURAL NETWORK FOR COMPUTERISED DETERMINATION AND QUANTIFICATION OF STAIN

DS Sahota1, AE James2, CJ Haines1, AMZ Chang1

The Department of Obstetrics and Gynaecology1 and Laboratory Animal Services Centre2, The Chinese University of Hong Kong, Prince of Wales Hospital, Hong Kong

A self organising map (SOM) is a self adapting neural network where the output reflects the closest possible response to a given input signal and is a form of nonparametric cluster analysis. This paper describes the use of the SOM to identify, detect and quantify atherosclerotic lesions present on the surface of the aorta of rabbits based on the different degree of stain absorbed. The portion of the aorta from the aortic valve to the coelic artery was obtained from 60 rabbits which had either been fed on cholesterol enriched diet, cholesterol plus oestrogen enriched diet or their standard diet for a 12 week period. These were cut longitudinally and immersed in a solution of Sudan III (4g in 70% alcohol) for fifteen minutes to stain any atherosclerotic lesions present on the surface of the artery. The specimens were then re-washed in saline and digitally scanned at a resolution of 300 dots per inch. Representative digital images were selected for determination of the number of different colour patterns. Pixels in the reference images were converted into their individual colour components and sequentially presented to an unsupervised SOM to allow grouping according to the major colour philla present. The unsupervised SOM categorised pattern variations of the 3 normalised inputs into one of 10 neural outputs, each of which was allocated a unique colour for retrospective display. The reference images were then reanalysed and the association between each of the 10 neuronal outputs and different immunostained regions of the captured specimen was determined. The resulting SOM produced was then incorporated into a customised computer program for automated processing of the remaining digitised images. The number of pixels allocated to each of these three major philla (background, stain, and unstained), representing surface area, was determined using a pixel counting method. The percentage of surface area of the atherosclerotic lesion (AR) was then calculated as the ratio of the lesion area (AL) to that of the artery surface area (AS). Quantitative assessment of the surface sudanophilia was possible in 58 arteries demonstrating that areas of tissue which had been stained can reliably be detected by a neural network. This approach offers the advantages of consistency and repeatability. The SOM once trained for a particular staining reagent may be used on other tissues of the same type provided that it is prepared in a similar fashion to that which the network was initially trained with.

Return to the Table of Contents

  

IMPROVING INTRAPARTUM SURVEILLANCE : (II) FETAL ECG MORPHOLOGY AND TIME INTERVALS

WJ van Wijngaarden, BK Strachan and DK James

Department of Obstetrics & Gynaecology, University of Nottingham, Queens Medical Centre, Nottingham NG7 2UH, UK

Objectives: Both time interval and morphological changes of the FECG have been seen to accompany fetal compromise during labour. No studies to date have evaluated a combination of these in addition to electronic fetal monitoring to improve on the detection of fetal compromise during labour. The first aim of this pilot study was to establish whether a combination of these FECG parameters in addition to electronic fetal monitoring (EFM) has a larger potential to improve on the detection of intrapartum fetal compromise than when used as a single addition to EFM. This exercise was repeated with different approaches to depict FECG abnormality. The second aim of this pilot study was to analyse the sequence of FECG parameters becoming abnormal during fetal compromise. Study Design: A retrospective analysis of 20 intrapartum FECG recordings of cases recruited in the Nottingham multicentre EFM versus EFM versus EFM plus FECG intrapartum monitoring trial. Results: The combination of time interval and morphological changes of the FECG in addition to EFM monitoring has the potential to detect more compromised fetuses without increasing intervention rates than when using EFM alone. The contribution of T/QRS changes in this potential improvement is smaller than that of time interval changes. Of all FECG parameters studied, time interval changes are the first and the most common to become abnormal during fetal compromise. Conclusion: The detection of fetal compromise during labour could be improved by using a combination of both FECG time interval and morphology changes. Earlier detection of fetal compromise may be possible by applying an increased index of suspicion when a cumulative calculation of the PR time interval - fetal heart rate relationship (ratio index) becomes abnormal even when other parameters are normal. Larger studies are required to evaluate these findings.

Return to the Table of Contents

CHOICES WITH CHILDBIRTH - AN OBSTETRIC CAL PROGRAM

A Vacca

Department of Obstetrics & Gynaecology, The University of Queensland, Australia

For some time medical educators have been exploring new and more effective methods of teaching essential clinical skills and of maintaining competency levels of doctors. Choices with Childbirth (CWCB), is an interactive computer assisted training program that has been developed to address some of these issues in relation to the teaching of Progress of Labour and Vacuum Extraction and to provide childbirth assistants with a means of acquiring information and clinical skills through a self-directed and independent learning process. Some of the design features of the program include: Essential information sections incorporating a wide range of multimedia Graduated study programs relevant to defined medical and midwifery user categories Interactive case studies designed to test the user’s knowledge and clinical skills An index, comprehensive bibliography and list of contents linked directly to the appropriate information sections A section designed for the information of mothers, childbirth educators and neonatal paediatricians and nurses, and which also serves as a summary of vacuum extraction Integration of all sections of the program with simple navigational tools and instructions, and guidelines on how best to use the program. As separate sections of the program were completed they were evaluated by selected representatives from appropriate levels of undergraduate and postgraduate training and clinical practice. Suggestions for changes and improvements were incorporated into the program where possible to ensure that the final product met end-user requirements. The interactive design and ease of use of the Choices with Childbirth CAL program should enhance all medical and midwifery birth attendants’ knowledge and reasoning skills about progress of labour and vacuum extraction. For this reason, CWCB should be a useful adjunct to teaching programs in institutions responsible for the delivery of obstetric care and for undergraduate and postgraduate training.

Return to the Table of Contents

 

THE CTG TUTOR

WJ van Wijngaarden , J Gardosi, T Vanner, M Terrett

PRAM, Department of Obstetrics & Gynaecology, University Hospital, Nottingham, UK.

Interpretation of the CTG is an important aspect of intrapartum care but its training is often haphazard and poorly organised. It is also one of the main areas for litigation. To address this issue, a computer aided learning (CAL) program was written using ‘Authorware’ software. It consists of 4 components: A tutorial which gives a quick overview of the principal features of the CTG. A module for the description of a CTG screened in as a bitmap, together with a short clinical history. The CAL program teaches the assessment of the features of the CTG by categorisation according to a standard, hierarchical system. A set of management options are prompted on screen, of which the correct combination has to be selected by true/false answers. Follow-up screens illustrate the likely results of a particular decision. Case specific tutorials give feedback on the actual and the correct management in each case. This method allows simulation of clinical situations, where the trainee is faced with a set of circumstances together with a CTG which he/she has to assess to formulate a management plan. The program is being used by medical students, midwives and doctors and can also act as a refresher for senior doctors. It encourages a standard approach to CTG interpretation and emphasises its predominant role as a screening tool whose predictive value is limited and whose use is therefore relatively easy once a standard set of guidelines are used. It can be applied for training as well as testing; we believe that proficiency in the interpretation of CTGs should be a minimal requirement before being given responsibilities on labour ward.

Return to the Table of Contents

TELEMEDICINE IN HONG KONG AND MAINLAND CHINA

M Hjelm

Taskforce on Telemedicine, Faculty of Medicine, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong

Telemedicine continues to be introduced world wide as a tool of increasing importance to improving the quality of health care, locally and providing specialist services in remote areas. It has been estimated that as many as one third of all hospitals in rural areas of the United States now use audiovisual means for providing health care to their patients and for distance education and training. In Hong Kong, a start was made to introduce telemedicine in 1996. Since then several major telemedical conferences have been held, linking hospitals in Beijing with Hong Kong and other sites overseas. Several projects have also been established to explore the possibility of using audiovisual means for, (I) radiological consultations between hospitals, (ii) linking homes for the elderly with a district general hospital for consultative purpose, and (iii) transmitting electrophysiological signals between hospitals. In all instances ISDN lines are used at a bandwidth of 384K. In mainland China, telemedicine is now considered a central element in improving the quality of health care. The Chinese Ministry of Health has established a special organisation, the Golden Health Medical Network Co. Ltd, to support hospitals and provincial health authorities in implementing telemedicine. In parallel, there is rapid progress in improving the quality of the national communications network. It can be expected that communication by ISDN lines will be possible between major provincial cities within the next two years. Recently a network of specialist centres was inaugurated, which will serve to provide expert opinion in complicated clinical cases. There are also plans to use satellite communication to remote areas, where the ISDN network would be difficult to install. Applications of telemedicine will equally focus on education, training and consultations. Leading hospitals have indicated their strong interest in linking up with Hong Kong as well for consultations and teaching. The Chinese University of Hong Kong has therefore established links with two hospitals in Beijing and will soon set up a similar system with a teaching hospital in Xinjiang. These hospitals will be used as test sites for developing appropriate telemedical procedures and their evaluation in terms of applicability and cost effectiveness.

Return to the Table of Contents

DEVELOPMENT OF GYNAECOLOGICAL ENDOSCOPY AUDIT DATABASE

PM Yuen, DS Sahota, AMZ Chang

Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, Prince of Wales Hospital, Hong Kong

The rapid introduction and development of endoscopic surgery has lead to concern about the practice and safety of the procedures. In order to assess the range of endoscopic procedures and their associated morbidity, a territory wide audit is being conducted by the Hong Kong College of Obstetricians and Gynaecologists (HKCOG). All hospitals in Hong Kong, 10 private and 12 public, performing gynaecological endoscopic surgery (GES) from 1 January 1997 participated in the audit. The original audit form was first designed by the GES SubCommittee of the HKCOG to record commonly encountered diagnoses, procedures and complications. The form recorded the patient’s demographic details, clinical diagnoses, operative procedures performed, operative techniques and any associated complications. The form was so designed that it required only simple selection from defined lists, minimising the inconsistency due to terminology differences, illegibility of hand writing and spelling mistakes. A window based relational database program was constructed using Borland Delphi Version 2.0 to allow a non-medically trained clerk to enter the data. The visual data entry screen interfaces were designed to mimic that of the audit form to try to ensure accurate cross transfer of data entered on the audit form into the computer. Where appropriate visual lookup lists of defined items were used to facility easy data entry and to avoid typing errors. During the construction and design of the program, it was realised that the original audit form was deficient and required modification. To facilitate report generation, those data variables on the original form which required free text entry to describe the situation had to be transform into defined categories. To assess the accuracy of the data entry, 100 records within the database were randomly selected and independently cross checked and compared against the original data in the audit form. The completeness of the defined lists was further evaluated based on a preliminary data analysis of the first 1000 records in the database. Where appropriate, additional items were added accordingly. The program was originally constructed for purpose of the centralised territory wide audit. The user friendly format employed actually allows direct data entry without the need for data transfer from the audit form. It is envisaged that the program will be adopted by individual hospital for continuous auditing of these surgical procedures.

Return to the Table of Contents

 

RISK FACTORS OF ENDOMETRIAL CANCER IN PATIENTS WITH ABNORMAL UTERINE BLEEDING

Shi Wei, Gu Meijiao

Gynecology and Obstetrics Department, Tongji Hospital, Tongji Medical University, Wuhan 430030

Objective: To interpret the risk factors of endometrial cancer and improve the early diagnosis. Because abnormal uterine bleeding is a symptom of endometrial cancer, we analyzed the risk factors of endometrial cancer in the women with abnormal uterine bleeding not caused by pregnancy and endometrial cancer. At the same time, we compared the risk factors in postmenopausal and premenopausal patients to observe if risk factors are different in the two groups. We also established an endometrial cancer screening program depending on its risk factors, symptoms and examination. Methods: 58 women with endometrial cancer and 4 women with endometrial atypical hyperplasia were selected. 320 women with abnormal uterine bleeding because of different causes except for endometrial cancer and pregnancy were taken as controls. These 320 women included 208 in premenopause period and 112 in postmenopause period. The data were divided into three groups. Group A consisted of 320 controls and 62 patients, group B (postmenopausal group) 112 controls and 38 menopausal patients and group C (premenopausal group) 208 controls and 24 nonmenopausal patients. Univariate analyses and multivariate nonconditional logistic regression with backward elimination were made in the three groups. Combined these high risk factors, symptoms with medical examination, an endometrial cancer risk degree screening computer programme was designed according to Bayes conditional probability threro under Turbo c 2.0. This programme was tested with our data. Results: (1) The age of endometrial cancer patients was 32-75 years old (average 53.47). 38.7% patients with endometrial cancer were in the premenopausal period and the other was in the postmenopausal period. Four women with negative D&C results before operation (8.1%) were verified as having endometrial cancer. Six women (five in premenopausal period) who were treated because of other reasons were found to have endometrial cancer after operation. (2) The risk factors associated with endometrial cancer in all the women with abnormal uterine bleeding included diabetes, a history of primary or secondary infertility, a history of hypertention, combining other tumours (include 2 theca cell tumours, 1 granulasa cell tumour and 1 breast tumour) at the same time, weight index > 28, the age of menopause > 50, nulliparity and familial hypertention or heart disease. (3) The risk factors of endometrial cancer in the premenopausal and postmenopausal women with abnormal uterine bleeding were obviously different. The risk factors in the postmenopausal patients were diabetes, weight index > 28, the age of menopause > 50, a history of hypertention, a history of primary or secondary infertility and vaginal discharge. But in the premenopausal patients, the risk factors included a history of primary or secondary infertility, other tumours, nulliparity and weight index > 28. (4) Our computer screening programme included 17 factors and 43 selections. A gynecologist could input the data of a new patient with abnormal uterine bleeding in order when he run shi.exe under ucdos surrounding. We used our data to demonstrate this screening programme. The possibility rate of cancer was more than 50% in 69.3% patients with endometrial cancer and the possibility of non-cancer was more than 50% in 94. 4% controls. In ten patients with endometrial cancer who were not diagnosed correctly, there had seven patients (70%) with the possibility of cancer was more than 50% when judged by our programme. Conclusion: The risk factors of endometrial cancer in the patients with abnormal uterine bleeding are correlated with their estrogen level and endocrine state. The estrogen rise in a high level for a long time is the most important factor in the patients with endometrial cancer. Endocrine dysfunction also is a main factor in the production and development of endometrial cancer. We consider that the risk factors in postmenopausal and premenopausal women with abnormal uterine bleeding are different. It is related closely with some elderly physical diseases (diabetes and hypertention) in postmenopausal patients, so endometrial cancer may be connected with endocrine dysfunction in elderly women. It relates closely with infertility and nulliparity in premenopausal patients. So a young woman with abnormal uterine bleeding should also be focused on, especially when the woman has infertility or nulliparity.

Return to the Table of Contents

 

DATABASE ERRORS: IMPACT ON RESEARCH INFERENCE

VJ Roach1, LY Hin1, KB Ng2 and MS Rogers1

Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong1 and University Hospital, University of Malaysia2.

Modern Obstetric research relies upon the collection and analysis of large quantities of clinical data, accumulated over years and stored in the form of computerised databases. In many instructions, these data are entered prospectively. Increasingly, trends in obstetric publications and practice are influenced by information derived from such databases. Errors can limit the quality of the data and the output. The most important source of error is omission of data essential to subsequent analysis and interpretation. For example, failure to record age or weight can make analysis of obstetric outcomes meaningless. More subtly, failure to include exact glucose values in an oral glucose tolerance test limits the data to a positive or negative result, precluding comment on the validity of the test itself. We have analysed a large database collected at the University Hospital in Kuala Lumpur, Malaysia, looking specifically at the relationship between carbohydrate intolerance in pregnancy and pregnancy induced hypertension. We will discuss errors and omissions in the database which limited the analysis and significance of our findings. Strategies for reducing error, including forward planning, handling of missing, conflictions and out of range data will be presented.

Return to the Table of Contents

 

A PRACTICAL MICROCOMPUTER STATISTICS SOFTWARE FOR OBSTETRICS AND PERINATAL DATABASE

Cao Jigong, Cao Jian, Tian Wen, et al.

Yantai Yuhuangding Hospital, Shandong 264000

Since 1992, we have completed the development and application of a practical microcomputer statistics software for obstetrics and perinatal baby database by using Fox-Pro relational model database language, UCDOS platform and structured program design. Some clinical informations in the database included the turient (name, hospitalization number, age, nation, height, education degree, occupation, work unit, address, anamnesis, diagnosis, labor way, nationality, identity card number), husband (name, age, education, nation, nationality, identity card number), perinatal baby (birth date and time, sex, deformity, pregnant times, delivery times, pregnant weeks, birth weight, fetal number, height, upper limb length, lower limb length, transformation, death cause, autopsy, diagnosis, remarks, birth medicine certificate number), family (past genetic disease history), and others (record writer, midwife or operator, community classfication), a total of 45 items. The clinical informations in code were inputted. The operation way was simple, convenient, rapid and saving. Some major functional modules in this system involved master menu display, medical record entry, revision and deletion, recording of statistics, display of statistic results, printing chart, baby birth medical certificate print processing, query function and display, copy of chronologic record to disk, returning of disk data to rigid disk database, a total 25 modules. The comprehensive query function was perfect, the query items and amount were randomly selected according to the need and combination. Query methods and kinds were not limited. This system has been improving through 4 years’ usage and it has possessed some characteristics of stable performance, rapid operation, easy extension and modification, transplantation, full screen Chinese character display and so on. It also has a variety of functions, such as statistics, conversion, synthetic query and issue of medical certificate for baby birth. This system can be applied in monitoring statistics of perinatal baby in comformity with the scheme for monitoring the perinatal baby birth deficiency in China to satisfy the clinical statistic analysis, scientific research, family planning, mother and child health care, newborn registration management and others. The hardware requirements of the system are lower so that 386 model microcomputer and the compatible machine fit for the use in hospitals.

Return to the Table of Contents

 

A COMPUTERIZED CENTRAL FETAL MONITOR CONTRIBUTED FULL MONITORING UNDER REDUCED STAFF EFFORTS AND ECONOMISED RECORDING CHART

K Maeda, M Utsu, N Yamamoto, M Serizawa

Department of Obstetrics & Gynecology, Seirei Hamamtsu & Mikatahara Hospital, Hamamatsu, Japan

Our experts knowledge system created in 1980 is applied in this central fetal monitor. Input of the system is the receivers for 6 telemetry transmitters attached to hospitalized women in antepartum as well as intrapartum stages. The patients are free from the confinement in the bed or chair, and mobile in any situation. The patient acceptance is good. The radiowave is modulated by fetal heart Doppler signal and uterine contraction. The receivers are connected to common autocorrelation FHR units and uterine contraction amplifiers. Output of these devices are connected to a MO disk and computerized FHR analyzers installed in each monitor. There is no chart recorder in the system. The computer program analyses FHR, and produces high-pitched sound when any significant change appears. The staff do not need continuously to watch the chart record, but only need to look at the monitor screen after listening the alarm tone. The computer memory is quickly played back for one to 12 hrs, and the hard copy of screen display is printed on common paper. The staff is free from the chart paper problem, and expensive chart paper is economised. The MO disk memorizes the signals for more than 10.000 hrs. Patient ID is input by sliding the patient card without keyboard. Acceptance of nurses has been very good. The system is a customized model of TOITU MF-7300. Eight months after the introduction of the system, the monitorings were made for 1,500 times. All intrapartum cases were fully monitored, and in addition, many high risk antepartum patients were repeatedly or continuously monitored by the system. An abruptio and 4 fetal distress cases were detected by the monitor and received C-section. No perinatal death was recorded in the period. Full intrapartum and sufficient antepartum monitoring were achieved by the system without increasing the efforts of the staff. About 6000 patients are monitored when the system is expanded to the maximum 16 simultaneous monitoring. The number is enough for the full monitoring in average maternity hospitals, and full external fetal monitoring improved outcome in our experiences. No use of expensive recording chart, reduction of chart storage space, and easy retrieval of past monitoring by the MO disk are the other advantages of the system.

Return to the Table of Contents

  

OBSTETRIC RISK PREDICTION ON THE WORLD WIDE WEB

R Derom

University of Cambridge and Katholieke Universiteit Leuven, UK and Belgium

The Quality Assurance in Maternity Care (QAMC) project has produced a Web page to demonstrate the potential benefits in combining three elements: interactive access to information via the World Wide Web; statistical risk models and data mining techniques; large databases of medical information. The primary aim of the project is to explore artificial neural networks for obstetric risk prediction. The project’s initial motivation is: (1) to provide early warning for adverse pregnancy outcome; (2) to construct useful models of risk from large databases of routinely-collected perinatal information; (3) to address theoretical issues, e.g. the selection of good predictors. The QAMC Web page deals with risk of failure to progress in labour; uses the Scottish Morbidity Record (known as SMR2) as database (771571 singleton births occurring between 1980-91); allows users to enter a description of a case in terms of maternal age, number of children now living, maternal height, previous obstetrical history and current health state as represented by 30 ICD-9 codes. Section one of the Web site reports the prevalence rates of specific patient profiles. Section two shows the predicted risk of adverse outcome according to three different risk models: logistic regression, an ensemble of artificial neural networks and a look-up table smoothed with a Dirichlet prior. This Web-based system demonstrates a number of practical ways to "mine" large amounts of perinatal data, and also shows how clinicians world-wide can gain access to this information. Several interesting problems were encountered in developing the system: the absence of relevant information in the database (note that the SMR2 data was not gathered specifically for risk prediction, but as a means to provide quality audit), questions about data reliability (e.g., inconsistencies in case notes), the imprecision of terms used in ICD-9 and other diagnostic codes. Other important issues include the interpretability of the results, the ability to generalise from the data (since patient characteristics and medical care change over time and across countries or regions), and medico-legal aspects of risk prediction. In conclusion, the QAMC project has shown that it is feasible to develop an easily accessible risk prediction system trained on a large medical database. The quality of risk predictions provided by that system is limited, primarily, by the type of perinatal data that is recorded and the reliability of that information.

Return to the Table of Contents

  

THE OBGYN AND THE WORLD WIDE WEB

JW van der Slikke*

Hospital ‘de Heel’, Postbus 210, 1500 EE Zaandam, The Netherlands

The next decade the internet will have an exploding importance for the medical profession, thanks to E-mail and the World Wide Web. In this paper the impact on daily work will be explained and several websites for the professional in obstetrics and gynaecology will be discussed. Gynaecologists/obstetricians are faced daily with new technologies and the consequences of a rapid changing society. They perceive these changes and developments perhaps earlier than any other medical specialist, but always more intensely. At this moment 20% of the national associations/colleges of Obstetrics & Gynaecology already launched their web-site. It gives them the chance to communicate with their members in an easy and very effective way. Among the firsts was the Dutch Society of Obstetrics and Gynaecology with their website NVOG-net (http://www.nvog.nl/). A website that is not founded by a national society is OBGYN.net (http://www.obgyn.net/). It has developed from a professional discussion-list and is visited by more than a thousand fellow-obgyn’s a day from all over the globe. It became the world’s most important site for the OBGYN-professional. There are also special-interest websites (ultrasound, oncology, genetics). FERTI.net (http://www.ferti.net) is fully dedicated to the science and practice of assisted fertilization and human reproduction. Its primary aim is to stimulate information dissemination among professional workers, researchers and patients in this field. The impact of internet and the different characters of these and other websites will be discussed. * Also with the help of Bruce Speyer (OBGYN-net) and Han Schnek (FERTI-net)

Return to the Table of Contents

OBSTETRIC & GYNAECOLOGIC ULTRASOUND REPORTING, ANALYSIS AND ARCHIVING SYSTEM

J Walstab, F Scott, FY Chan

Department of Maternal Fetal Medicine, Mater Mothers’ Hospital, University of Queensland, Australia

We have developed a computer system for obstetric & gynaecologic ultrasound reporting, analysis and archiving based on Windows Filemaker-pro. The programme has the following features : Data entry by keyboard or by direct download from an ultrasound machine (aTL). Automatic report generation for the following types of scans: first trimester dating, second trimester dating and morphology screen, third trimester growth & wellbeing, genetic procedures (such as amniocentesis, chorionic villus biopsy), twin pregnancies, general gynaecology, gynaecological infertility. Automatic data calculations for percentile fetal size according to various formulae. Automatic graph generation to show fetal growth and umbilical arterial Doppler flow percentiles. Permanent and comprehensive database of all data including systematic analysis of fetal anomalies. Automatic generation of statistics for the practise (monthly or yearly), including generation of letter to initial referring doctors re pregnancy outcome. Interface with most other database systems. Being a window based programme, the programme is user-friendly, and has networking capabilities to allow data from various ultrasound machines to be stored and retrieved in several stations.

Return to the Table of Contents

FETAL MONITOR BASED ON ULTRASONIC WAVE DIRECT DIGITAL DETECTION

Y Yamakoshi

Faculty of Engineering, Gunma University, Japan

A goal of the fetal monitor may be quantitative multi parameter monitoring which relates with the fetal well being. For example, high accuracy beat to beat fetal heart rate monitoring and quantitative measurement of the fetal movement will give valuable information in fetal well being evaluation. In designing such a system, both spatially localized ultrasonic wave detection and signal processing optimized for each fetal well being parameter are absolutely required. In this paper, fetal monitor based on ultrasonic wave direct digital detection is proposed. The raw ultrasonic wave signals are directly A/D converted and are processed in a digital way inside the specially designed LSI (Large Scaled Integrated circuit). Both internal tissue displacement for the fetal movement and the fetal heart rate are estimated simultaneously. Features of the system are; 1) Since many spatially localized ultrasonic wave detectors are equipped, quantitative multi parameter fetal monitoring is executed easily. 2) Since all the signal processing is done in a digital way, fetal heart rate is measured with high accuracy due to higher signal to noise ratio of the system. The basic experiments are carried out and the usefulness of the system for quantitative multi parameter fetal monitoring is demonstrated.

Return to the Table of Contents

A WINDOWS BASED SYSTEM FOR TREND ANALYSIS OF INTRAPARTUM CTGS

GJ Colenbrander1, HP van Geijn2

1Department of Clinical Physics and Engineering. 2Department of Obstetrics and Gynecology, Academic Hospital Vrije Universiteit, P.O. Box 7057, 1007 MB, Amsterdam, The Netherlands.

Since 1991 the department of Obstetrics and Gynecology uses a network based DOS application for the acquisition, monitoring, storage and archiving of cardiotocograms (CTG’s). Every cardiotocograph is on-line digitized, displayed on a composite screen and stored on disk. This abstract describes its structure, functionality, recently developed analysis tools and the change to a Windows based system. Hardware: The application runs on a hospital wide network (UTP, Ethernet) with a number of fileservers. The local fileserver for this application is a Pentium 90 PC with 32 Mb RAM connected to a 2 Gb SCSI hard drive for data storage and a MO-drive for archiving. A DAT tape unit is attached for backups. The workstations vary from 486 PC to Pentium Pro PC’s. Notebooks are used for the mobile acquisition systems in the nursery wards. Software: In 1996 we changed from Banyan Vines as network operating system to Windows NT 4. The network clients run Windows 95 but attempts are made to run Windows NT 4. The main databases can be accessed simultaneously by other applications (like a report or an analysis system). Analysis tools: To analyse trends in intrapartum CTG’s we needed a flexible Windows based system that fulfills the following conditions: suitable user interface, compatibility with the present CTG acquisition system, selection of max. 6 hours CTG, identification of fetal behavioural states, possibility for data selection and import/export of the data, statistical analysis and integration with other applications. The MOSOS<CA> system will be presented that supports most of these functions. Our aim is to extend this system to a complete research toolbox for analysis of intrapartum CTG’s.

Return to the Table of Contents

 

RECENT APPROACHES FOR NEW FETAL HEART RATE MONITORING TRANSDUCERS

J Morgenstern, D Westhues

Heinrich Heine Universitat Dusseldorf, Germany

During the last decade, signal processing power has significantly increased, leading to new computational power also in fetal heart rate processing technology using chaos theory and other new signal processing approaches. On the other hand, for the same time there were no new basic approaches in research and development of new kinds transducers suitable for the use with the method of cardiotocography. It should be obvious that the more power engaged in research of the front end of a signal processing chain the less computational power is needed to extract the true signal. Depending on the algorithm used in fetal monitors, fetal heart rate traces are more or less integrated. The amount of signal integration is directly related to an artificial closing of trace gaps. Signal loss is mainly related to a misalignment of the ultrasound beam caused by fetal movements and/or uterine contractions. Among some new approaches of which none has yet reached a market level, is an ultrasound transducer to overcome those gaps in fetal heart rate traces during fetal heart rate monitoring. In close comparison to well-known radar systems an ultrasound transducer with a small movable piezocrystal is designed. The piezoelement in the transducer is held by a cardan joint, therefore being turnable up to 45 in all directions. The crystal can be hand-adjusted as well as steered by two micro-motors while the position is controlled by two rotary encoders. The aim was to compensate for any ultrasound-beam-misalignment by keeping it mechanically adjusted on the fetal heart. The system specifications are defined by controlling the transducer with a simple computer controlled feedback tracking algorithm in a water tank. Other attempts aim at an integration of the two existing ultrasound-heart-rate- and the strain-gage-labor transducer into one housing. Instead of using several small piezo elements, one single large disc is used with a convex, movable lens. The lens itself also serves as a sensig element for the strain gage. In a fully digital fetal heart rate monitor prototype, a I-III piezo composite is used to overcome unwanted shere-mode vibrations. Challenging for material engineering is here the large diameter of 60 mm. Incoporation of telemetry systems into all above mentioned transducers seem to be desirable to increase comfort of patients.

Return to the Table of Contents

FETAL HEART RATE TRANSFERRING SYSTEM BY APPLYING GAME MACHINE

M Hogaki1, Y Takcuchi2, J Morgenstern3

1 School of medicine, Teikyo Univ., Tokyo, 2 GE YMS, Tokyo, Japan 3 Department of Obstetrics & Gynecology, Heinlich-Heine University, Dusseldorf, Germany

Low cost data transferring system for simultaneous monitoring of the multiple patients of Fetal Heart Rate (FHR) is demonstrated by applying a modified game machine with a simple low cost interface box installed for every FHR meter. Instrumentation; Each box from the instrument is serially connected by a line to a modified cheap game machine working as a net work computer just collecting and buffering the incoming data, combined with another low cost personal computer with CPU less than 30 MHz more than 5 times slower than those of high end machines, with relatively large memory (32-144 MB) and hard disc (2-4 GB), omitting keyboard, just applicable patients cards. The total system is designed for the simultaneous monitoring of 4 patients determined as a optimal velocity for the applied game machine, being restricted with those of the low cost CPU. In cases of less number of patients, the total system can be automatically activated by a simple procedure of the key operation as a recognition of working FHR meter. Clinical evaluation; In this system, there no process of specific network forming connection just affording automatic recognition of bed number, without any complicated procedure of activating many diagnostic softwares, except for the detecting the incoming data size in accordance with the patient number. Another characteristics of the system is the elimination of the sophisticated diagnostic softwares, by affording the simplified automatic detection of the FHR baseline. In case of further low cost application of the total system, conventional FHR meter can be set up by using game machine, by adding simplified auto correlation softwares.

Return to the Table of Contents

DEVELOPMENT OF AN ALGORITHM FOR BASELINE ANALYSIS OF NEONATAL HEART RATE

GJ Colenbrander1, M Van Gelder2, RJA Peters1, JIP De Vries2, HP Van Geijn2

1Department of Clinical Physics and Engineering. 2Department of Obstetrics and Gynecology, Academic Hospital Vrije Universiteit, P.O. Box 7057, 1007 MB, Amsterdam, The Netherlands

Introduction: Pioneer work of Dawes et al. resulted in the equipment SYS8000 for baseline analysis1. Mantel et al. modified it applying the concept of cycling behavioural states2. Arduini developed a Windows based system for baseline analysis3. In neonatology statistical indices are available to quantify heart rate (HR) variability, but a true baseline analysis is lacking4. However, to understand the continuity in baseline HR and its variability from fetal to neonatal life, it is of importance to use the same baseline analysis. Study design: In a longitudinal study on the development of ischemic brain lesions in high-risk pregnancies 20 fetuses were monitored for heart rate and motility during 1 hour. During neonatal life heart rate (ECG), motility, SpO2 and breathing frequency were monitored. Methods: The baseline algorithm according to Mantel et al. Was modified to comply with the neonatal heart parameters. After preprocessing and artcfact detection/rejection a baseline filter was applied to the data. This filtering process consisted of (1) analysis of the distribution of RR intervals, (2) interpolation of missing data, (3) and a low-pass filter. Finally accelerations and decelerations were detected. Implementation: A DOS multichannel acquisition system POLYTM was used to gather, process and analyse the data. This system has adjustable sample rates per sufficient analysis tools. To apply the baseline filter the data size of HR signals was largely reduced (resampling at fs=0.4 Hz (2.5s)). Conclusion: Baseline filtering is a helpful tool to assess fetal as well as neonatal HR and HRV. Future research will focus on the relation between the heart rate, motility and SpO2. References: Dawes GS, Moulden M, Redman CWG. Criteria for the design of fetal heart rate analysis systems. Int J Biomed Comput, (1990), 25:287-294. Mantel R, Ververs IAP, Colenbrander GJ, Van Geijn HP. Automated antepartum baseline FHR determination and detection of accelerations and decelerations. In: Van Geijn HP, Copray FJA (eds.). A Critical Appraisal of Fetal Surveillance, Elsevier, Amsterdam, 1994, 333-348. Arduini D, Rizzo G, Piana G, Bonalumi A, Brambilla P, Romanini C. Computerized analysis of fetal heart rate: I. Description of the system (2CTG). J Matern Fetal Invest (1993) 3:159-163. Van Ravenswaaij-Arts CMA, Hopman JCW, Kollee LAA, van Amen JPL, Stoelinga GBA, van Geijn HP. Influences on heart rate variability in spontaneuously breathing preterm infants. Earl Hum Dev (1991) 27:187-205.

Return to the Table of Contents

 

MAGNETIC RESONANCE TOMOGRAPHY IN GYNECOLOGICAL PRACTICE

L Erofeeva

Moscow Organon Information Centre, Research Centre of Obstetrics, Gynecology and Perinatology, Moscow, Russia

The recent years are characterized by the wider application of new technologies in gynecological practice. Magnetic resonance tomography (MRT)is assumed to be one of the promising method of diagnosis. The objectives of the study included the assessment of diagnostical value of MRT as compared to other methods. The study was carried out with the apparatus Fleaxart (Toshiba, the Netherlands). One-hundred seventy-eight patients were examined (76 ovarian cysts, 30 tubo-ovarian masses of inflammatory character, 51 uterine myomas, 21 endometriosis). The use of MRT allowed to reveal the sites of malignization of the masses, blood circulation disorders in the uterine myomas, adhesions. Diagnostical value of the method was 92%. The obtained results allow to conclude that MRT is a promising diagnostical method in gynecological practice.

Return to the Table of Contents

PERINATAL DATABASE OF THE DEPARTMENT OF OBSTETRICS AND GYNECOLOGY, UNIVERSITY OF THE PHILIPPINES

MR Festin, L dela Cruz, CP Mangubat, AF Poblete, MLO Otayza.

University of the Philippines Manila, Department of Obstetrics and Gynecology, Philippine General Hospital, Taft Avenue, Manila

Background. Since 1993, the department has been using a perinatal database of all deliveries, for use in the departmental audit and reports. Objectives. To describe the perinatal database used by the department, to list the main independent and dependent variables listed, and to list the advantages and problems encountered in the use of the database. Methods. The database uses DbaseIII+ for data collection and Epi-Info V. 6.01 for the analysis. Residents and fellows take turns in the data entry tasks. Quarterly perinatal audit reports are made based on the database output. Results: Since 1993, 20,000 patients have been included in the database. 100 dependent variables are listed, including gestation age, maternal risk factors, antenatal care, labor and delivery patterns, and birth data. The primary independent variable in fetal outcome, specifically live births, stillbirths, early and late neonatal deaths. Problems include various levels of computer literacy of the staff, changes in the database structure, computer viruses and repairs. However, these are far outweighed by the advantages, including faster analysis of the data, standardization of reporting, and availability of the database for further study. The database is now being tried in the home institutions of the graduated trainees in their own audits. Conclusions: A perinatal database provides ample data access for audits, research and information. Basic computer training may be incorporated in the training program of residents and fellows to allow utilization of such information.

Return to the Table of Contents

A QUALITY ASSESSMENT OF DELIVERY REGISTRATION DATABASE

Terzic N, Marinkovic J, Kocev N, Labovic I

Institute of Public Health - Republic of Montenegro, Podgorica, Yugoslavia

To follow work of the gynecological departments, in order to assure efficient health protection during the pregnancy, delivery and puerperium or planning on this area, it is very important that data referring to the registration of deliveries to be complete and accurate. The aim of this paper is examination quality of data that are gathered from computerized database of delivery registrations. Database employees all deliveries done in Montenegro since 1992. This database for statistical data processing concerning deliveries contains essential data set about data on anamnesis and partus, course of puerperium, infant data. There have been analyzed 8069 deliveries in 1995. Quality assessment for all data fields from database had been done on 1% systematic subsample (27) of sample (2566) of the registration deliveries happened in 1995 in Podgorica, by comparison analysis of data in medical documentation and record fields. Results of the examination shows data concerning complication in pregnancy and puerperium are the least reliable because it has not been noticed any complication. This is also confirmed with the examination of data on the hole sample by descriptive analysis (97%) without pregnancy complications, only 4 women with puerperium complications), and this is not reality. The fewest reliable data about newborn infants are those concerning the state on discharge (15% with errors) and the pathology of newborns. The most accurate data are the age of mother, number of the hospital-days and infant weights. The quality of these data does not show the objective state in the health protection of mother and child, but reliable maintaining of these data would enable advancement of gynecology-obstetrics offices, planning and prediction on this area.

Return to the Table of Contents

COMPUTER ANALYSIS IN MANAGEMENT OF FAMILY PLANNING SERVICE IN UDMOURDIA REPUBLIC

K Serebrennikova, L Erofeeva

Izhevsk Medical Instidute, Izhevsk, Russian Federation

To improve the main demographic indeces in the Republic of Udmourdia a computer analysis was made of the mortality rate, birth rate of rural and urban popuation, abortion rate. A decrease in birth rate was proved: in 1994 as compared to 1984 by 1 mln. Within the framework of the programme Health of Women and Family Planning in the Republic there were elaborated and introduced into practice family planning service with the organizational and commercial justification of its functions, training programme for the specialists. The introduction of family planning programme contributed to the decrease of abortions rate from 126.3 per 1000 women of fertile age in 1991 to 85.3 in 1995.

Return to the Table of Contents

MULTI-PURPOSE DATABASE FOR CHORE OF MICROBIOLOGIST

AV Golubev, SG Smirnov and LV Posiseyeva

Ivanovo Scientific-Research Institute of maternity and childhood named after V.N.Gorodkov, Russia.

In the microbiological laboratory of Ivanovo Scientific-research Institute of maternity and childhood named after V.N.Gorodkov is introduced and is successfully use by itself design and developed by the personnel a laboratory specialized applied system "Heuristic laboratory". This software package oriented on the operation in the ambience Windows-95, Windows NT4 (version on Visual FoxPro 5.0), or in the ambience Windows for Workgroups 3.11 or WIN-OS/2 WARP (version on FoxPro 2.6 for Windows) and being, practically, complex of automatic working places of microbiologists allowing produce an individual evaluation clinical and analytical information for each patient and/or groups of patients on the whole collection of available data. Provided by possibility of automatic syntheses of the varied documents, find using in the chore a labs.; conduct of operative databanks on the clinic-analytical parameters of all registered patients, on presence of reassets and spent material etc. Provided by possibility of working a system in local network. System is organized in nineteen databases, bound by subscripted fields, screens of entering information and service subroutines. Operative work with the complex is relieve by presence multiayered menu and operative context help system. Each of main sections is intended for working with certain information and consists of database specifications, indexes, programmes, input and output forms. Last were created in the calculation on the unprofessional user and correspond to developed and introduced by us standard forms of answers, conclusions, etc. In purposes of achievement greatly reliable and qualitative using a programme complex , in the majority of screens of entering information is aplying original by itself designed algorithm, allowing ensure a high velocity and software guarantee correctness and confidentiality of introductory data. The whole information system as a whole executes main functions databases: sorting, searching, processing and issue information in suitable for the unprofessional user type. Provided by possibility of sheduling the samples any degrees of difficulty on any kit of signs, coding and cryptooperation information, layered password access, etc.

Return to the Table of Contents

FORMALIZED FEATURES OF DYNAMIC MICROBIAL ASSOCIATIONS IN OBSTETRIC PRACTICE

AV Golubev, SG Smirnov and LV Posiseyeva

Ivanovo Scientific-Research Institute of maternity and childhood named after V.N.Gorodkov, Russia.

Take aim the work was an association study Enterobacteriaceae and Candida, existing on mucous generic ways of woman with the different degree of nongestating pregnancy, and practicability of using of formalizes factors for the feature of clinical value of these associations. Object of study take microbial associations, chosen beside 183 woman. Including beside two groups nonpregnant and beside 6-ty of groups pregnant with the different degree pathological deflections during pregnancy - from the packed rate before miscarriages at different terms. Microbiological studies included a study of microflora in the three biotops: in the back code a vagina, cervical channel, upper respiratory fetters. Microbial background was research dynamically with staking out contents in serum shelters infection factors. All types of material was produce simultaneously. For the determination infection status is applying diagnostic center Quantum-II and technologies of company ABBOTT. Were identify and studied 1884 cultures 50 types, from them bacterias 1527 (81.05%, 33 types) and fungus 357 (18.95%, 17 types). From the microbiological standpoint biochemical heterogeneity (Hk), biochemical activity (Ha) and, particularly, specific biochemical heterogeneity (Hu) in greater depth result of pregnancy to conditions external (for microflora) ambiances, i.e. to the condition of macroorganism of pregnant women on the background general "unifying" nature of action of adjustment factors to gestation period. Proposed method of evaluation of results of microflora studies with the determination developed by us parameters opens one more way for the achievement of more full description speakers of vital activity an microassociations with provision for the concrete substrata of its haunting and, in the prospect will allow do one more step on way of making the mathematical microbial process models. Results of conducting studies and developed criterion are to promote an improvement of microbiological diagnostics of diseases in obstetric practice.

Return to the Table of Contents

A NEW ALGORITHM FOR RELIABLE DETECTION OF THE FETAL ELECTROCARDIOGRAM FROM THE MATERNAL ABDOMEN

RM Lewinsky, A Mizrahi, A Fux, D Lange, GF Inbar, G Ohel

Department Obstetrics & Gynecology, Bnai Zion Medical Center and Faculty of Electrical Engineering, Technion Institute of Technology, Haifa, ISRAEL

Background: Multiple attempts have been made in the past to obtain a fetal ECG signal from the maternal abdomen. These attempts which required complex equipment yielded varying degrees of success and were unsuitable for routine clinical use. We present a new, easy to operate system, which invariably detects the fetal ECG, making its clinical use a reality. Material and Methods: One pair of electrodes is applied to the maternal abdomen at the extremes of its midline longitudinal axis. In addition, a conventional Doppler-ultrasound transducer is applied for the detection of fetal cardiac signal. Both analog signals are extracted from the commercial FHR monitor and sampled by an IBM-compatible computer at a rate of 500 Hz for further analysis. The two dominant peaks on the fetal Doppler cardiogram, S4 and S1 are detected for the generation of a search window for the fetal QRS complex which is located between them. A single-beat model is applied to generate an average template of the maternal and fetal ECG complexes. Both templates together with an auto regressive model of noise are implemented in the reconstruction of the pure fetal ECG signal. Results: The system was first tried on 16 patients in labor with a fetal scalp ECG being used as a reliable reference. All fetal ECG complexes were detected from the abdominal signal. Twenty four pregnant women with gestational age of 26-41 weeks were then studied antenatally. In the presence of an audible Doppler signal, the system was capable of extracting a clear fetal ECG from the maternal abdomen in all cases. There was no need for re-positioning of the abdominal electrodes. Again, all fetal ECG complexes, even those overlapping maternal complexes were detected. Conclusions: Antenatal recording of the fetal ECG from the maternal abdomen is feasible with a high degree of reliability without need for complex or expensive equipment. No special expertise is needed to operate this system. The pure fetal ECG signal can be used for the diagnosis of fetal arrhythmia, and for the evaluation of fetal well-being by power spectral analysis of FHR variability and by waveform analysis.

Return to the Table of Contents

THE EMPLOYMENT OF THE METHOD FOR MANIFESTATION VARIED DATAS FOR THE PURPOSE TO GIVE AN APPRECIATION OF FUNCTIONAL TESTS IN THE PROCESS OF REABILITATION OF WOMEN AFTER HYSTERECTOMY

GB Dikke, VA Volovodenko

Tomsk Research Institute of Curortology and Phisiotherapy, Tomsk Politechnical University, Russia

77.1% women who came through an operation of hysterectomy have symptoms of neuro-vegetative, psycho-emotional and metabol -endocrinological disturbance. Functional aggravation at vegetative nervous system (VNS) is plaing a leading part of it’s beginning. The object of our research was the giving an appreciation of functional state of VNS with helping of method for manifestation varied datas. 70 women at the age 28-42 years were examinated after hysterectomy. Examination of VNS consisted of determination of initial data of vegetative tone, ortoclinostatics and cold tests. The results were worked up with the help of pockets of statistic softwares "statgraf" and the method for manifestation of varied datas at the computer IBM PS with included construction three-dimensional conception for varied originals which their present datas as well as their total combinations. Our method made it possible to select groups of datas which express itself in special feature of images of numerical characteristic’s vectors. It made it possible to trace the process of formation of the groups on the level of integrated presentation of everyone of research organisms. The using of present method showed the dynamics of functional state of VNS after reabilitation of women who came through an operation of hysterectomy and show the normalisation of vegetative reactions and improvement of reserve means of constitution.

Return to the Table of Contents

COMPUTERISED ASSESSMENT OF VASCULAR DIAMETER DURING TREATMENT WITH GONADOTROPIN RELEASING HORMONE AGONISTS AND HORMONE REPLACEMENT THERAPY

SF Yim, CJ Haines, TK Lau, DS Sahota, TKH Chung, AMZ Chang

Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, Prince of Wales Hospital, Hong Kong

This study investigated the effects of GnRHa administration and "add back’ therapy on vascular reactivity my measuring artery changes using a computerised digital micrometer. Measurements of endothelium dependent and endothelium independent vascular reactivity were compared in 2 groups of women treated with a GnRHa for 6 months. One group received oestrogen/progestogen "add back" therapy during the second three months of GnRHa treatment. Vascular reactivity was examined using ultrasound measurements of changes in brachial artery diameter. Vascular reactivity was examined by directly capturing the ultrasound image of the brachial artery using a video framegrabber. A computerised software digital micrometer was designed and calibrated to measure changes in the brachial artery diameter. This differs from previous studies which used the calipers on the ultrasound machine which reduced the precision in measurement of artery diameter changes. Endothelium dependent changes were assessed during reactive hyperaemia, whilst endothelium independent changes were measured following the administration of glyceryl trinitrate sublingual spray. Treatment with the GnRHa resulted in a significant inhibition of endothelium dependent relaxation in the brachial artery. Endothelium dependent relaxation was significantly greater in the group which received ‘add back’ therapy (14.6%) than the group treated with GnRHa alone (8.6%) (p<0.01). There were no significant endothelium independent changes in either group. These results suggest that the administration of ‘add back’ therapy has a protective effect on vascular function in women using GnRHa agonists.

Return to the Table of Contents 

COMPUTER-ASSISTED ANALYSIS OF NORMAL AND SUBNORMAL SEMEN SAMPLES

EPL Loong, TTY Chiu, CJ Haines

Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, Prince of Wales Hospital, Hong Kong

Introduction Routine semen analysis usually forms part of the initial investigation of male factor infertility. Although it is an important diagnostic test, conventional light microscopy lacks objectivity and is subject to individual variations when examined by different investigators. The assessment of sperm motility is particularly affected since the accuracy is influenced by the technical conditions and the skill of the observer. Computer-assisted analysis can offer an objective assessment of various sperm parameters. In this study, we compared the results of normal and subnormal semen using the Hobson Tracker System (U.K.). Materials and Methods A total of 38 semen samples were collected during the routine procedures for IVF of which 21 were normal and 17 were subnormal samples. The results of the initial analysis were classified in accordance with WHO criteria (1992). Computer-assisted analysis was then performed on each sample using the Hobson Tracker System. The percentage of sperm motility and various sperm kinematic parameters i.e. curvilinear velocity (VCL), straight line velocity (VSL), linearity (LIN), amplitude of lateral head displacement (ALH) and beat cross frequency were studied. At least 100 tracks were recorded before the completion of each analysis. The variations in motility assessed by conventional and computerized analysis were expressed by their respective coefficient of variation. Comparison of other semen parameters obtained by the Hobson Tracker were analyzed by the Student’s t test. Results Table 1. Assessment of sperm motility by conventional microscopy and by the Hobson Tracker System Method Normal semen Subnormal semen (n=10) (n=10) Conventional microscopy 63.9 ± 7.5 (11.7%*) 28.0 ± 7.2 (25%*) Hobson Tracker System 84.3 ± 5.2 (6.2%*) 36.0 ± 2.2 (6.1%*) * Values are coefficient of variation expressed as percentages. Table 2. Assessment of sperm characteristics by the Hobson Tracker System Parameters Normal semen Subnormal semen (n=21) (n=17) Curvilinear velocity (VCL) 112.5 ± 14.0 92.5 ± 18.1* Average path velocity (VAP) 43.9 ± 6.9 40.6 ± 5.7 Straight line velocity (VSL) 28.8 ± 7.3 23.7 ± 7.7* Linearity (LIN) 26.2 ± 6.8 24.8 ± 8.3 Beat cross frequency (BCF) 14.5 ± 2.9 8.2 ± 4.3* Amplitude of lateral head 6.9 ± 1.6 6.7 ± 2.3 displacement (ALH) * p < 0.05 as compared to normal samples. Conclusion The assessment of motility using the computer-assisted analyzer (Hobson Tracker System) appears to be more reliable and shows a much smaller variability as compared to conventional microscopic analysis. Furthermore, computer-assisted analysis can give a more comprehensive assessment of other parameters which may have prognostic value to subsequent IVF outcome.