Related Experiment Video
Updated: Aug 29, 2025
09:57
Author Spotlight: Advancing Pediatric Epilepsy Surgery in Children Through Novel Biomarkers and Enhanced Localization
Published on: September 20, 2024
2.8K
Machine learning algorithm performance evaluation in structural magnetic resonance imaging-based classification of
Ruhai Dou1, Weijia Gao2, Qingmin Meng3
1Department of Radiology, Shandong First Medical University and Shandong Academy of Medical Sciences, Taian, China.
Frontiers in Computational Neuroscience
|September 9, 2022
Summary
Machine learning accurately aids pediatric bipolar disorder diagnosis by analyzing brain MRI scans. This approach improves diagnostic accuracy, identifying key brain regions for better classification of pediatric bipolar disorder.
Area of Science:
- Neuroscience
- Psychiatry
- Computer Science
Background:
- Clinical diagnosis of pediatric bipolar disorder (PBD) faces challenges with potential misdiagnosis.
- Machine learning (ML) offers promising tools for improving the classification accuracy of bipolar disorder (BD).
Purpose of the Study:
- To evaluate the efficacy of ML methods in classifying pediatric bipolar disorder (PBD) using neuroimaging data.
- To identify key neuroimaging features indicative of PBD.
Main Methods:
- Extracted brain cortical thickness and subcortical volumes from MRI data of 33 PBD-I patients and 19 healthy controls (HCs).
- Applied dimensionality reduction (Lasso, f_classif) and six ML classifiers (LR, SVM, Random Forest, Naïve Bayes, k-NN, AdaBoost).
- Trained and tested classifiers to assess diagnostic performance.
Main Results:
- Logistic Regression (LR) achieved 84.19% accuracy, and Support Vector Machine (SVM) achieved 82.80% accuracy.
- Identified the right middle temporal gyrus and bilateral pallidum as crucial features for PBD classification.
- Results align with known structural and functional brain abnormalities in PBD.
Conclusions:
- ML-based analysis of MRI data provides a valuable tool for the computer-aided diagnosis of pediatric bipolar disorder.
- The identified brain regions highlight potential biomarkers for PBD.
- This study advances the application of ML in psychiatric diagnostics.
