AI enabled surgery: enabling da Vinci Xi robot with live cancerous tumor detection
Deep learning has proved successful in Computer Aided Detection in interpreting ultrasound images, CT scans, identifying COVID infections, identifying tumors from ultrasound, Computed Tomography (CT) scans for humans and for animals. Currently, only experienced surgeons can identify tumors in patients with kidney cancer using ultrasound and Indocyanine Green (ICG) with Fluorescence Imaging which may come with error. Therefore, this project proposes applications of deep learning in detecting cancerous tissue inside patients via laparoscopic camera on da Vinci Xi surgical robots. The proposed algorithm can help the surgeons to detect cancerous tumors from fatty tissue and non-cancerous tissue with 84% accuracy during the surgery which is extremely beneficial to ensure all the cancerous tumors are removed. The process is carried out via object detection techniques which draws bounding boxes and shows the probability for that region to be cancerous tissue, non cancerous tissue or fatty tissue, which is the primary goal of the project. The project compares between optimized AlexNet, VGG-16, YOLOv3, YOLOv4 to work out the best algorithm with tuned hyperparameter to detect cancerous tissue during surgery. Analysing images, the final mAP for object detection was 0.974 and for classification, the accuracy was 0.84.