Cross-Modal Fine-Tuning of 3D Convolutional Foundation Models for ADHD Classification with Low-Rank Adaptation

2025-11-08

AI Summary

Addressing the challenges of high symptom heterogeneity and substantial cross-disorder overlap in neuroimaging-based diagnosis of pediatric ADHD, this study introduces, for the first time, a cross-modal transfer learning paradigm that adapts a CT-pretrained 3D convolutional foundation model to MRI. We propose a novel 3D low-rank adaptation (LoRA) method, decomposing 3D convolutional kernels into 2D low-rank updates—enabling parameter-efficient fine-tuning while preserving architectural integrity. Our approach requires only 1.64 million trainable parameters—over 113× fewer than full fine-tuning—and achieves 71.9% accuracy and an AUC of 0.716 under five-fold cross-validation, setting a new state-of-the-art. This work establishes the first CT→MRI cross-modal neuroimaging diagnostic benchmark and provides a scalable, data-efficient paradigm for AI-assisted diagnosis of psychiatric disorders characterized by small sample sizes and high clinical heterogeneity.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Multi-modal Vision
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Early diagnosis of attention-deficit/hyperactivity disorder (ADHD) in children plays a crucial role in improving outcomes in education and mental health. Diagnosing ADHD using neuroimaging data, however, remains challenging due to heterogeneous presentations and overlapping symptoms with other conditions. To address this, we propose a novel parameter-efficient transfer learning approach that adapts a large-scale 3D convolutional foundation model, pre-trained on CT images, to an MRI-based ADHD classification task. Our method introduces Low-Rank Adaptation (LoRA) in 3D by factorizing 3D convolutional kernels into 2D low-rank updates, dramatically reducing trainable parameters while achieving superior performance. In a five-fold cross-validated evaluation on a public diffusion MRI database, our 3D LoRA fine-tuning strategy achieved state-of-the-art results, with one model variant reaching 71.9% accuracy and another attaining an AUC of 0.716. Both variants use only 1.64 million trainable parameters (over 113x fewer than a fully fine-tuned foundation model). Our results represent one of the first successful cross-modal (CT-to-MRI) adaptations of a foundation model in neuroimaging, establishing a new benchmark for ADHD classification while greatly improving efficiency.

Problem

• Classifying ADHD using neuroimaging data with heterogeneous presentations
• Adapting CT-trained 3D models to MRI-based ADHD classification efficiently
• Reducing trainable parameters while maintaining high classification performance

Innovation

• Adapts 3D convolutional foundation model via cross-modal fine-tuning
• Introduces Low-Rank Adaptation by factorizing 3D convolutional kernels
• Dramatically reduces trainable parameters while maintaining performance