Machine Learning and Graph Signal Processing Applied to Healthcare: A Review

Area of Science:

• Signal processing
• Graph theory
• Machine learning
• Health sciences

Background:

• Signal processing is crucial for interpreting everyday signals.
• Graph theory extends signal processing to non-Euclidean domains for time-varying signals.
• Machine learning is widely applied in pattern recognition, including health sciences.

Purpose of the Study:

• To identify and analyze literature on machine learning applied to graph signal processing in health sciences.
• To understand the current state and identify research gaps in this emerging area.

Main Methods:

• A systematic literature search was conducted across four major databases: Science Direct, IEEE Xplore, ACM, and MDPI.
• Specific search strings were used to identify relevant papers.
• A total of 45 papers published from 2015 onwards were included in the analysis.

Main Results:

• The application of machine learning to graph signal processing in health sciences is an emerging research area, with the first publications appearing in 2015.
• Analysis revealed a need for enhanced clinical interpretability of results beyond mere performance metrics.
• Identified research gaps include the exploration of novel transforms and the creation of new public datasets.

Conclusions:

• Machine learning applied to graph signal processing is a growing field within health sciences.
• Future research should focus on improving clinical relevance and developing new methodologies and accessible datasets.
• Addressing these gaps will further advance the practical application of these techniques in healthcare.