Endometriosis detection and localization in laparoscopic gynecology

Abstract Endometriosis is a common gynecologic condition typically treated via laparoscopic surgery. Its visual versatility makes it hard to identify for non-specialized physicians and challenging classify or localize computer-aided analysis. In this work, we take first step in the direction of localized endometriosis recognition gynecology videos using region-based deep neural networks Faster R-CNN Mask R-CNN. We particular use further develop publicly available data transfer learning detection models according distinctive lesion characteristics. Subsequently, evaluate performance impact different augmentation techniques, including selected geometrical transformations, specular reflection removal as well region tracking across video frames. Finally, attention given creating reasonable segmentation training, validation testing. The best performing result surprisingly achieved by randomly applying simple cropping combined with rotation, resulting mean average precision 32.4% at 50-95% intersection over union overlap (64.2% 50% overlap).