Bleeding contour detection for craniotomy

Computer-vision based analysis of the neurosurgical scene - A systematic review

Introduction

With increasing use of robotic surgical adjuncts, artificial intelligence and augmented reality in neurosurgery, the automated analysis of digital images and videos acquired over various procedures becomes a subject of increased interest. While several computer vision (CV) methods have been developed and implemented for analyzing surgical scenes, few studies have been dedicated to neurosurgery.

Research question

In this work, we present a systematic literature review focusing on CV methodologies specifically applied to the analysis of neurosurgical procedures based on intra-operative images and videos. Additionally, we provide recommendations for the future developments of CV models in neurosurgery.

Material and methods

We conducted a systematic literature search in multiple databases until January 17, 2023, including Web of Science, PubMed, IEEE Xplore, Embase, and SpringerLink.

Results

We identified 17 studies employing CV algorithms on neurosurgical videos/images. The most common applications of CV were tool and neuroanatomical structure detection or characterization, and to a lesser extent, surgical workflow analysis. Convolutional neural networks (CNN) were the most frequently utilized architecture for CV models (65%), demonstrating superior performances in tool detection and segmentation. In particular, mask recurrent-CNN manifested most robust performance outcomes across different modalities.

Discussion and conclusion

Our systematic review demonstrates that CV models have been reported that can effectively detect and differentiate tools, surgical phases, neuroanatomical structures, as well as critical events in complex neurosurgical scenes with accuracies above 95%. Automated tool recognition contributes to objective characterization and assessment of surgical performance, with potential applications in neurosurgical training and intra-operative safety management.

Autonomous aspirating robot for removing saliva blood mixed liquid in oral surgery

Saliva blood mixed liquid (SBML) appears in oral surgery, such as scaling and root planning, and it affects surgical vision and causes discomfort to the patient. However, removing SBML, i.e. frequent aspiration of the mixed liquid, is a routine task involving heavy workload and interruption of oral surgery. Therefore, it is valuable to alternate the manual mode by autonomous robotic technique. The robotic system is designed consisting of an RGB-D camera, a manipulator, a disposable oral aspirator. An algorithm is developed for detection of SBML. Path planning method is also addressed for the distal end of the aspirator. A workflow for removing SBML is presented. 95% of the area of the SBML in the oral cavity was removed after liquid aspiration among a group of ten SBML aspiration experiments. This study provides the first result of the autonomous aspirating robot (AAR) for removing SBML in oral surgery, demonstrating that SBML can be removed by the autonomous robot, freeing stomatology surgeon from tedious work.

Deep learning-based classification and segmentation for scalpels

PURPOSE

Scalpels are typical tools used for cutting in surgery, and the surgical tray is one of the locations where the scalpel is present during surgery. However, there is no known method for the classification and segmentation of multiple types of scalpels. This paper presents a dataset of multiple types of scalpels and a classification and segmentation method that can be applied as a first step for validating segmentation of scalpels and further applications can include identifying scalpels from other tools in different clinical scenarios.

METHODS

The proposed scalpel dataset contains 6400 images with labeled information of 10 types of scalpels, and a classification and segmentation model for multiple types of scalpels is obtained by training the dataset based on Mask R-CNN. The article concludes with an analysis and evaluation of the network performance, verifying the feasibility of the work.

RESULTS

A multi-type scalpel dataset was established, and the classification and segmentation models of multi-type scalpel were obtained by training the Mask R-CNN. The average accuracy and average recall reached 94.19% and 96.61%, respectively, in the classification task and 93.30% and 95.14%, respectively, in the segmentation task.

CONCLUSION

The first scalpel dataset is created covering multiple types of scalpels. And the classification and segmentation of multiple types of scalpels are realized for the first time. This study achieves the classification and segmentation of scalpels in a surgical tray scene, providing a potential solution for scalpel recognition, localization and tracking.