Knowledge-Informed Machine Learning for Cancer Diagnosis and Prognosis: A Review

Area of Science:

• Oncology
• Artificial Intelligence
• Bioinformatics

Background:

• Machine learning (ML) aids cancer diagnosis and prognosis through complex data analysis.
• ML models face limitations including small sample sizes, high-dimensional data, patient heterogeneity, and interpretability issues.
• Integrating biomedical knowledge into ML models can enhance accuracy, robustness, and interpretability.

Purpose of the Study:

• To review state-of-the-art machine learning studies that fuse biomedical knowledge and data for cancer research.
• To explore knowledge-informed machine learning (KIML) applications in cancer diagnosis and prognosis.
• To discuss future directions for KIML in advancing cancer research and healthcare automation.

Main Methods:

• Review of current literature on knowledge-informed machine learning in oncology.
• Analysis of diverse knowledge representation forms and integration strategies.
• Examination of concrete examples of KIML in cancer diagnosis and prognosis.

Main Results:

• KIML demonstrates potential to overcome ML limitations in cancer research.
• Successful integration of biomedical knowledge improves ML model performance.
• Diverse strategies exist for representing and integrating knowledge into ML pipelines.

Conclusions:

• KIML is a promising approach to advance cancer diagnosis and prognosis.
• Further research into KIML can enhance healthcare automation in oncology.
• An evolving online resource is available to support KIML research in cancer.