VP62.16: Artificial intelligence in the detection of rectosigmoid deep endometriosis

the aim of this study was to compare the accuracy of seven classical machine learning models of ultrasound (US) soft markers in raising suspicion of bowel involvement using a dataset with previously published results. The data driven strategy has been to split randomly the complete dataset in two different datasets. The training dataset and the test dataset with a 67% and 33% of the original cases respectively. All models have been trained on the training dataset and the predictions have been evaluated using the test dataset. The election of the best model will be based on the accuracy demonstrated on the test dataset. The variables using in all the models were: age; presence of US signs of uterine adenomyosis; presence of an endometrioma; adhesions of the ovary to the uterus; presence of “kissing ovaries”; absence of sliding sign. All models have been trained using CARET package in R with ten repeated 10-fold cross-validation. For every model a confusion matrix is presented. Accuracy, Sensitivity, Specificity, VPP and VPN have been calculated using a 50% threshold. All new cases in the test dataset with an estimated probability greater than 0.5 have been classified as presence of intestinal involvement. A summary results table shows the performance of the seven models. The best model is the Neural Net without significant difference with the others. The accuracy of US soft markers in raising suspicion of rectosigmoid endometriosis using AI models showed similar result in comparison with the logistic model. This study was partly supported by Fondazione di Sardegna grant F74I19001010007. VP62.16: Table 1. k-nearest neighbors algorithm (k-NN) Support Vector Machine (SVM)