Multitask learning in minimally invasive surgical vision: A review

Area of Science:

• Computer Vision
• Surgical Robotics
• Machine Learning

Background:

• Minimally invasive surgery (MIS) offers benefits but increases complexity for surgical teams.
• Advancements in machine learning and computer vision are crucial for developing autonomous MIS.
• Analyzing MIS videos presents challenges due to the complexity of surgical scenes and actions.

Purpose of the Study:

• To provide a narrative overview of state-of-the-art multitask learning (MTL) systems for MIS video analysis.
• To discuss the benefits and limitations of current MTL approaches in MIS.
• To analyze the literature on MTL applications in MIS, identifying trends and future research directions.

Main Methods:

• Review of existing literature on MTL applied to surgical videos from MIS.
• Analysis of various application fields utilizing MTL in MIS.
• Discussion of large model integration within MTL frameworks for MIS.

Main Results:

• MTL effectively addresses challenges in surgical scene and action understanding by leveraging related tasks.
• MTL improves performance and generalization in analyzing complex MIS video data.
• Identified trends and new research directions in MTL for MIS.

Conclusions:

• MTL is a promising paradigm for enhancing the understanding of MIS videos.
• MTL systems have the potential to significantly contribute to the development of more autonomous and capable MIS systems.
• Further research into MTL, particularly with large models, is warranted for advancing MIS technology.