A survey on deep learning for surgical planning

Abstract

Deep learning-based surgical planning is currently a promising area of research. Unlike the traditional way of surgical planning through computer systems that assist physicians, deep learning methods enhance the robustness and accuracy of surgical planning systems through a data-driven approach with the great advantage it possesses in processing images. In this work, we have put together a review that presents the application of deep learning in three separate aspects of surgical planning systems, namely, surgical scene understanding, surgical scene reconstruction and automated assessment of surgical skills. This article covers multiple scopes, such as semantic segmentation, depth estimation, SLAM systems, etc. We present the relevance of these three applications for the current surgical and medical field and show how the problem can be solved by the techniques in the mentioned topics. We hope that this work will link emerging research results in the field of deep learning and surgical planning, and provide guidance to future researchers using deep learning techniques for surgical planning when it comes to understanding feasible approaches to related problems.