Diagnostic value of structural and diffusion imaging measures in schizophrenia

Area of Science:

• Neuroimaging
• Machine Learning
• Psychiatry

Background:

• Distinguishing schizophrenia from healthy controls is a significant challenge in clinical neuroscience.
• Previous research has utilized structural and functional MRI for classification, with varying success.
• Integrating multimodal MRI data, including diffusion MRI, may enhance diagnostic accuracy.

Purpose of the Study:

• To evaluate the efficacy of machine learning algorithms, specifically random forest (RF) and support vector machine (SVM), in discriminating schizophrenia patients from healthy controls.
• To assess the performance of these models using a combination of structural and diffusion MRI (dMRI) features.
• To compare the classification performance of RF and SVM against chance levels.

Main Methods:

• Utilized 504 features (volume, fractional anisotropy, trace) from 184 brain regions derived from structural and dMRI.
• Employed a dataset of 47 schizophrenia patients and 23 healthy controls, balanced using Synthetic Minority Oversampling Technique (SMOTE).
• Performed leave-one-out cross-validation and random permutation testing (100 permutations) to assess model robustness and significance.

Main Results:

• Both RF and SVM models demonstrated significantly superior classification performance compared to chance.
• RF achieved 87.6% sensitivity and 95.9% specificity; SVM achieved 89.5% sensitivity and 94.5% specificity.
• These results indicate high accuracy in discriminating schizophrenia using the combined MRI features and machine learning.

Conclusions:

• Machine learning models, particularly RF and SVM, effectively discriminate schizophrenia patients using multimodal MRI data (structural and diffusion).
• The integration of volume and diffusion measures significantly enhances classification performance.
• Further validation through larger, independent studies is warranted to confirm these findings.