MALINI

MALINI provides MATLAB-based implementations of 18 diverse machine learning classifiers and a consensus classifier for feature extraction and diagnostic classification of resting-state functional MRI (rs-fMRI) connectivity data.

Key Features:

• Diverse Classifiers: Implements 18 machine learning classifiers based on different algorithmic principles for classification of rs-fMRI connectivity features.
• Consensus Classifier: Aggregates predictions across all 18 classifiers to mitigate overfitting and improve generalizability across heterogeneous samples.
• Feature Importance Analysis: Integrates feature importance scores from each classifier to assess the discriminative power of functional connectivity features.
• Robustness Across Datasets: Evaluates connectivity patterns for algorithm-independence and resilience to differences in participant age and acquisition site.
• Cross-sample Validation: Supports training and validation of classifiers on samples with the same diagnosis that vary in age or acquisition site to assess generalizability.

Scientific Applications:

• Autism Spectrum Disorder (ASD): Applied to rs-fMRI datasets totaling N=988 to evaluate classifier generalizability and identify diagnostic connectivity patterns.
• Attention Deficit Hyperactivity Disorder (ADHD): Applied to rs-fMRI datasets totaling N=930 to assess classification performance across heterogeneous samples.
• Post-Traumatic Stress Disorder (PTSD): Applied to rs-fMRI datasets totaling N=87 to investigate diagnostic connectivity signatures in a clinical cohort.
• Alzheimer's Disease (AD): Applied to rs-fMRI datasets totaling N=132 to evaluate discriminative connectivity features for diagnostic classification.

Methodology:

Classifiers are trained and validated on samples with the same diagnosis but varying in age or acquisition site; a consensus classifier combines predictions across the 18 classifiers, and feature importance scores from each classifier are integrated to evaluate discriminative connectivity features.