Surgical action detection based on path aggregation adaptive spatial network

Surgeon action detection plays a crucial role in computer-assisted surgery. However, due to the problems of non-rigid instrument deformation, occlusion, and less available contextual information in the surgeon action detection, these factors lead to the problem that the average accuracy of the current detection of surgical motion is very low, which needs to be solved urgently. Therefore, inspired by the application of convolutional neural networks (CNNs) can express features through learning and success in medical image detection tasks, we developed a path aggregation adaptive spatial feature pyramid network (PAAS-FPN), which combines bottom-up path enhancement and an adaptive spatial fusion mechanism. Path enhancement can use the shallow feature information of images for upward transmission. The adaptive spatial feature fusion network adds spatial granularity between deep and shallow features. In this study, the improved method was experimentally verified on the ESAD dataset and surgeon instrument detection dataset. The proposed detection method achieved the highest detection accuracy in several experiments, thereby confirming its effectiveness in surgeon action detection.