Liquid white box model as an explainable AI for surgery.

Area of Science:

• Medical Artificial Intelligence
• Surgical Informatics
• Machine Learning in Healthcare

Background:

• Real-time surgical data analysis is crucial for enhancing surgeon feedback, learning, and performance.
• Data-driven systems promise safer, more standardized surgeries and accelerated training.
• Artificial intelligence (AI) can address limitations in human and classical computing for efficient information processing.

Purpose of the Study:

• To develop explainable AI models for surgical task and skill classification.
• To provide transparent explanations for AI-driven surgical decision-making.
• To investigate the use of liquid time constant models for improved performance under constraints.

Main Methods:

• Development of two distinct AI models: one for surgical task classification and another for skill classification.
• Implementation of explainability features to elucidate model decision processes.
• Investigation of liquid time constant models for enhanced performance and interpretability.

Main Results:

• Successfully created AI models capable of classifying surgical tasks and skills.
• Demonstrated the models' ability to predict and explain surgical decisions.
• Showcased the effectiveness of liquid time constant models in constrained environments.

Conclusions:

• Explainable AI models can significantly improve real-time surgical data understanding and application.
• Transparent AI decision-making is essential for robust model development and adoption in surgery.
• Liquid time constant models offer a promising approach for developing effective and interpretable AI in surgical contexts.