Abstract:
Segmentation of surgical pictures and videos is a critical task that involves the precise separation of surgical significant regions at the pixel level. In recent years, the medical industry has witnessed a significant surge in interest towards this particular subject, mostly owing to its superior precision in contrast to object detection techniques. The ARAS robotics group has conducted a comprehensive range of activities pertaining to the analysis and assessment of surgical proficiency. In this regard, ARAS-Farabi software has been developed, which transparently performs automatic and intelligent analysis and evaluation of surgical skills. This thesis expands on the earlier research conducted by the ARAS Robotics Group in the fields of image-video data processing during cataract surgery and the crucial capsulorhexis procedure. It also establishes a framework for the more precise extraction of movement information from the surgical key regions. One of the challenges encountered in this thesis is to the diminutive dimensions of the surgical instrument and the absence of a distinct boundary in numerous corneal images. The utilization of the image-video dataset has been prevalent owing to the widespread presence of cameras in all operating rooms, as well as the absence of the requirement for costly equipment, in contrast to sensor data. Under these circumstances, a comprehensive annotated segmentation dataset with applicability in diverse domains such as medicine and computer vision has been developed in consultation with the surgeons at Farabi Hospital. In this research, State-of-the-art image segmentation structures have been implemented on the prepared dataset. Additionally, an accomplishment of this thesis is the video segmentation of capsulorhexis surgery. Lastly, The extraction of movement information from segmented videos of capsulorhexis surgery has been undertaken in order enable the potential of surgical skill analysis. The findings derived from this study, owing to their high pixelwise precision, have the potential to be utilized in the development of ARAS-Farabi software, which can be employed for the analysis and assessment of surgical proficiency.