339 General discussion, conclusions, and future perspectives promising start. Machine learning (ML) can be implemented in the field of surgical skills analysis for a multitude of tasks which were until now, performed manually. In line with the review by Andras et al. we believe AI will have a growing presence in the operating room45 AI will not only assist in the assessment of technical surgical skills but also in Non-Technical skills (NTS) assessment.46 Further development of organ recognition in combination with the use of imaging modalities as MRI and CTscans could eventually result in the development of surgical devices which can provide intraoperative navigation using augmented reality based on MRI or CT-scan data. This could ensure the surgeon removes the organ or tumour with clean surgical margins and the maximum respect for the surrounding tissues. It is even conceivable the implementation of AI will lead to the development of a surgical robot that can operate with less or no human interference using tissue and landmark recognition.47,48 The relation between physician performance and patients’ postoperative outcomes To date, little research has aimed to correlate technical skills assessment using surgical videos to the patient’s postoperative outcome15,42,44,49,50. The difficulty of this field of research is identification of the factors involved in prediction of postoperative outcome. In this thesis, multiple techniques of surgical video analysis have been investigated to identify factors involved in the origins of postoperative incontinence and erectile dysfunction. If these analyses of pre-recorded surgical videos could be fully automated using DL and AI objective assessment could be implemented, leading to further insights into the surgeon’s performance. The group of Hung et al. has performed multiple studies into this new field of research.42,44,51 These initial studies using machine learning and automated performance metrics have resulted in new insights into the effects of surgeon’s performance on post-operative outcome.42,44 Based on the current assessment of the surgical videos by expert surgeons the length of the urethra and the quality of the neuro-vascular bundles could respectively affect incontinence and erectile dysfunction. The relation of the per-operatively measured length of the urethra and postoperative continence has been proven in this thesis. The possibility of urethral length measurements in previously recorded videos (surgical videos of older surgeries) could revolutionize the field of surgery. If DL and AI can be used to recognize structures in the surgical video, the length of the residual urethra stump could be measured accurately and automatically. If this is done in previously recorded surgical videos, the next step will be to implement automated measurements during surgery, enabling the surgeon to maximize the length of the residual urethra and thus
RkJQdWJsaXNoZXIy MTk4NDMw