Performing and Evaluation of Deep Learning Models for Uterus Detection on Soft-tissue Cadavers in Laparoscopic Gynecology

Nowadays, with the current technological forces that have been shaping our bright future, one of these is Computer Vision. And this statement true across various matters, including laparoscopic gynecology, where computer-aided procedures for object recognition could offer surgeons opportunity to ease up on on-going surgeries and/or practice their surgical skills offline surgeries. However, most previous work has retrospective and focused methodology from a computational viewpoint minimal datasets showing how Vision can be utilized surgery. The main purpose paper not just evaluate state-of-the-art detection models uterus detection, but also emphasize clinical application via collaboration between peopleware which important in further development adoption technology, leading improved outcomes Laparoscopic Gynecology. Two experiment phases conducted. Phase#1 applied 8 different Deep Learning were tested dataset, obtained 40 public YouTube videos Gynecologic Surgery. In order prove new technology before performing patients, due ethics human experimentation, extensive testing soft-tissue cadavers used, hence Phase#2 performed best first phase serving real-time streaming feed during 4 cadaver surgeries, theoretically approach as they are closest humans terms shape structure. Four models, pre-trained COCO 2017 Dataset TensorFlow Model Zoo: CenterNet; EfficientDet; SSD; Faster R-CNN; plus YOLOv4 Darknet Framework, along YOLOv4, YOLOv5 YOLOv7 Pytorch scrutinized here. inference time (in FPS: Frame Per Second), F1-score AP (Average Precision) used evaluation metrics. results exhibited all 3 YOLOs PyTorch outperformed effectiveness metrics, great speed suitable Lastly, by-product useful contribution work, annotated dataset both live