Neural Network Analysis of AnalSphincter Repair

PURPOSE: Prediction of success after anterior sphincter repairfor incontinence is difficult. Standard multivariateanalysis techniques have only 75 to 80 percent accuracy.Artificial intelligence, including artificial neural networks,has been used in the analysis of complex clinical data andhas proved to be successful in predicting the outcome ofother surgical procedures. Using a neural network algorithm,we have assessed the probability of success afteranterior sphincter repair. METHODS: Prospective anorectalphysiology data of 72 patients undergoing anterior sphincterrepair was collected between 1995 and 1999. Completedata sets of 75 percent of the series were used to train anartificial neural network; the remaining 25 percent wereused for data validation. The output was continence grading,ranging from 0 to 4 (worse to continent). RESULTS: Theoutcome at 3, 6, and 12 months postoperatively was obtainedand assessed. The best correlation between actualdata value and artificial neural network value was found at12 months (r = 0.931; P = 0.0001). Clear correlations alsowere found at three months (r = 0.898; P = 0.0001) and sixmonths (r = 0.742; P = 0.002). Results of applying a net todetails excluding pudendal nerve latency were poor. CONCLUSIONS:Artificial neural networks are more accurate (93percent correlation) than standard statistics (75 percent)when applied to the prediction of outcome after anteriorsphincter repair. This assessment also confirms the usefulnessof pudendal latency in the prediction of anteriorsphincter repair outcome. The results obtained highlightthe obvious usefulness of artificial neural networks, whichcould now be used in a prospective evaluation for applicationof the technique.