Effects of a medical expert system on differential diagnosis of renal masses: a prospective study

Cross-domain probabilistic inference in a clinical decision support system: examples for dermatology and rheumatology

INTRODUCTION

Maintaining a large diagnostic knowledge base (KB) is a demanding task for any person or organization. Future clinical decision support system (CDSS) may rely on multiple, smaller and more focused KBs developed and maintained at different locations that work together seamlessly. A cross-domain inference tool has great clinical import and utility.

METHODS

We developed a modified multi-membership Bayes formulation to facilitate the cross-domain probabilistic inferencing among KBs with overlapping diseases. Two KBs developed for evaluation were non-infectious generalized blistering diseases (GBD) and autoimmune diseases (AID). After the KBs were finalized, they were evaluated separately for validity.

RESULT

Ten cases from medical journal case reports were used to evaluate this "cross-domain" inference across the two KBs. The resultant non-error rate (NER) was 90%, and the average of probabilities assigned to the correct diagnosis (AVP) was 64.8% for cross-domain consultations.

CONCLUSION

A novel formulation is now available to deal with problems occurring in a clinical diagnostic decision support system with multi-domain KBs. The utilization of this formulation will help in the development of more integrated KBs with greater focused knowledge domains.

Neural network modeling for surgical decisions on traumatic brain injury patients

Computerized medical decision support systems have been a major research topic in recent years. Intelligent computer programs were implemented to aid physicians and other medical professionals in making difficult medical decisions. This report compares three different mathematical models for building a traumatic brain injury (TBI) medical decision support system (MDSS). These models were developed based on a large TBI patient database. This MDSS accepts a set of patient data such as the types of skull fracture, Glasgow Coma Scale (GCS), episode of convulsion and return the chance that a neurosurgeon would recommend an open-skull surgery for this patient. The three mathematical models described in this report including a logistic regression model, a multi-layer perceptron (MLP) neural network and a radial-basis-function (RBF) neural network. From the 12,640 patients selected from the database. A randomly drawn 9480 cases were used as the training group to develop/train our models. The other 3160 cases were in the validation group which we used to evaluate the performance of these models. We used sensitivity, specificity, areas under receiver-operating characteristics (ROC) curve and calibration curves as the indicator of how accurate these models are in predicting a neurosurgeon's decision on open-skull surgery. The results showed that, assuming equal importance of sensitivity and specificity, the logistic regression model had a (sensitivity, specificity) of (73%, 68%), compared to (80%, 80%) from the RBF model and (88%, 80%) from the MLP model. The resultant areas under ROC curve for logistic regression, RBF and MLP neural networks are 0.761, 0.880 and 0.897, respectively (P < 0.05). Among these models, the logistic regression has noticeably poorer calibration. This study demonstrated the feasibility of applying neural networks as the mechanism for TBI decision support systems based on clinical databases. The results also suggest that neural networks may be a better solution for complex, non-linear medical decision support systems than conventional statistical techniques such as logistic regression.

Evaluation of a decision-support system for preoperative staging of prostate cancer

The usefulness and effectiveness of a decision-support system for preoperative staging of prostate cancers (PCES) were evaluated. The study population consisted of 43 consecutive patients with the preoperative diagnosis of prostate cancer who underwent surgical operation. Results obtained using the PCES were compared with staging by four urology attending physicians and five urology residents. The effect of PCES consultation on the physicians' staging of prostate cancer was also evaluated. To confirm the usefulness of the clinical findings of prostate-specific antigen, prostate-specific antigen density, prostate volume, and abnormal Gleason score in the PCES, their receiver operating characteristic (ROC) curves for diagnosis of advanced prostate cancer were plotted. The values of the areas under the curves were 0.772, 0.800, 0.531, and 0.752. The stage of prostate cancer was correctly determined by the PCES for 38 of the 43 patients, yielding 88.4% preoperative diagnostic accuracy. The PCES was significantly more accurate than two of the attending physicians and all residents. PCES consultation improved the residents' staging accuracy to approximately that of the attending physicians. The effect of PCES consultation on the residents' staging was significantly (p < 0.001) greater than the effect on the physicians' staging. The PCES may be useful in the preoperative staging of prostate cancers, especially during residency. The system's accuracy in determining the stage of advanced prostate cancer may make it possible to avoid unneccesary surgical operations.