Related Experiment Video
Updated: Jun 16, 2026
07:35
Selecting Multiple Biomarker Subsets with Similarly Effective Binary Classification Performances
Published on: October 11, 2018
Good things come in threes: evaluating clinical utility of machine learning-derived clusters
Daniil Lisik1,2, Jip W T M De Kok3,4, Nicolás Bermúdez Barón2
1Department of Public Health and Clinical Medicine, The OLIN and Sunderby Research Unit, Umeå University, Umeå, Sweden.
JAMIA Open
|June 15, 2026
Summary
This study proposes a framework to ensure machine learning (ML) clusters are clinically useful. It addresses limitations in current ML applications for disease subtyping, aiming for better real-world healthcare integration.
Area of Science:
- Medical Informatics
- Computational Biology
- Clinical Data Science
Background:
- Machine learning (ML) cluster analysis is widely used for medical condition subtyping.
- Increasing data and algorithmic advances boost ML model use in healthcare.
- Lack of guidelines hinders practical implementation due to methodological and reporting issues.
Purpose of the Study:
- To propose a general framework for assessing ML-derived cluster clinical utility.
- To provide guidelines for evaluating ML-generated subgroups in medicine.
Main Methods:
- The framework focuses on three key domains for assessing ML-derived clusters.
- Domains include clinical relevance, stability, generalizability, and ease of identification.
Main Results:
- The proposed framework offers a structured approach to evaluate ML-generated subgroups.
- It aims to bridge the gap between ML methodology and clinical application.
Conclusions:
- Implementing this framework can enhance the clinical utility of ML-derived clusters.
- Standardized assessment is crucial for reliable ML application in disease subtyping.