Quantitative evaluation of camera-based 3D reconstruction in laparoscopy: A Review

The quantitative evaluation of 3D reconstruction methods might support the successful improvement and extension of computer assisted surgery in regard of applications as augmented reality and (autonomous) robot-assisted interventions. There is a selection of around eleven evaluation metrics from which the root mean squared error, the mean absolute error and the standard deviation are the most used. For this work a literature review was done with the objective to give a statement about promising 3D reconstruction methods regarding accuracy. Therefore, quantitative evaluation results of 44 articles from 2015 to 2022 were analysed. The review includes the reconstruction methods stereoscopy, Shape-from-Motion, Shape-from-Shading, Simultaneous Localization and Mapping, artificial intelligence-based approaches (machine-learning), structured light, multicamera and trinocular approaches and the so-called Smart Trocar®. It can be identified that machine-learning approaches in combination with stereo images and stereoscopy in combination with structured light deliver the best results with values in submillimetre range, thus seem to be promising for future research.