Sex-specific ADHD pattern discovery: a hybrid machine learning approach using functional and social-clinical data

The complex neuro-developmental disorder known as attention deficit hyperactivity disorder(ADHD) is characterized by impulsivity, hyperactivity, and persistent inattention. Because of overlapping symptoms and sex-specific variability, early and accurate diagnosis is still a clinical challenge. For better diagnostic accuracy, a hybrid machine learning framework in this study that combines behavioral and sociological characteristics with functional connectome data. Using logistic regression as the meta-learner and Extra Trees and Gradient boosting classifiers as base learners, a stacked ensemble learning model was used. To identify potential sex specific ADHD signatures, the model was trained and assessed using a stratified multimodal dataset that focused on both ADHD diagnosis and sex classification. An 82.71% classification accuracy for ADHD and a 75.30% classification accuracy for sex were obtained from evaluation on test set, indicating the model’s ability to deconstruct intricate cross-modal interactions. This work support the use of ensemble-based approaches in sex-aware, data-driven, neurodevelopmental diagnostics and highlight their effectiveness in identifying clinically meaningful patterns.