Related Experiment Video
Updated: Jun 17, 2026
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
15.7K
Machine learning in small sample neuroimaging studies: Novel measures for schizophrenia analysis.
Carmen Jimenez-Mesa1, Javier Ramirez1, Zhenghui Yi2
1Department of Signal Theory, Telematics and Communications, Andalusian Research Institute in Data Science and Computational Intelligence (DaSCI), University of Granada, Granada, Spain.
Human Brain Mapping
|March 28, 2024
Summary
This study introduces a novel computer-aided diagnosis (CAD) system using brain sulcal patterns to distinguish schizophrenia patients from healthy individuals. The system demonstrates effective classification even with limited data, highlighting key differentiating brain regions.
Area of Science:
- Neuroimaging
- Artificial Intelligence
- Computational Psychiatry
Background:
- Computer-aided diagnosis (CAD) systems integrate imaging and AI for clinical support and biological pattern investigation.
- Existing CAD systems often lack comprehensive evaluation frameworks, especially for novel feature types and small datasets.
Purpose of the Study:
- To develop and evaluate a CAD system for classifying schizophrenia using structural brain imaging-derived sulcal patterns.
- To demonstrate a comprehensive evaluation methodology for CAD systems with limited empirical guidance and small sample sizes.
Main Methods:
- Extracted sulcal features from the entire cerebral cortex of 58 schizophrenia patients and 56 healthy controls.
- Applied sequential statistical, machine learning, and deep learning techniques for classification.
- Utilized explainable AI (XAI) methods (LIME, SHAP) to identify feature relevance and dimensionality reduction for optimization.
Main Results:
- Successfully differentiated schizophrenia patients from controls based on sulcal patterns.
- Identified specific differentiating sulcal patterns in temporal and precentral areas, and the collateral fissure.
- Validated the benefits of dimensionality reduction and advanced validation methods for small sample performance optimization.
Conclusions:
- Sulcal patterns from the entire cerebral cortex can serve as effective biomarkers for schizophrenia detection using CAD systems.
- The proposed comprehensive evaluation framework is suitable for developing CAD systems with novel features and limited data.
- Explainable AI techniques are valuable for understanding feature importance in neuroimaging-based CAD systems.
