AD-Reasoning: Multimodal Guideline-Guided Reasoning for Alzheimer's Disease Diagnosis

2026-03-25

AI Summary

This work proposes AD-Reasoning, a novel framework for Alzheimer’s disease (AD) diagnosis that integrates structural MRI with six categories of clinical data to align with the National Institute on Aging–Alzheimer’s Association (NIA-AA) guidelines. Addressing the lack of transparency and guideline adherence in existing models, AD-Reasoning employs modality-specific encoders and bidirectional cross-attention for multimodal fusion, coupled with a rule-driven verifier that generates structured diagnoses and interpretable reasoning paths compliant with clinical standards. The framework further introduces a verifiable reward mechanism to guide reinforcement fine-tuning, ensuring output format consistency, comprehensive evidence coverage, and alignment between reasoning and diagnostic decisions. A large-scale multimodal question-answering dataset, AD-MultiSense, is also released. Experiments demonstrate that the proposed method significantly outperforms current baselines in both diagnostic accuracy and reasoning transparency.

Technology Category

• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

Alzheimer's disease (AD) diagnosis requires integrating neuroimaging with heterogeneous clinical evidence and reasoning under established criteria, yet most multimodal models remain opaque and weakly guideline-aligned. We present AD-Reasoning, a multimodal framework that couples structural MRI with six clinical modalities and a rule-based verifier to generate structured, NIA-AA-consistent diagnoses. AD-Reasoning combines modality-specific encoders, bidirectional cross-attention fusion, and reinforcement fine-tuning with verifiable rewards that enforce output format, guideline evidence coverage, and reasoning--decision consistency. We also release AD-MultiSense, a 10,378-visit multimodal QA dataset with guideline-validated rationales built from ADNI/AIBL. On AD-MultiSense, AD-Reasoning achieves state-of-the-art diagnostic accuracy and produces structured rationales that improve transparency over recent baselines, while providing transparent rationales.

Problem

• Alzheimer's disease
• multimodal diagnosis
• clinical guidelines
• transparent reasoning
• heterogeneous evidence

Innovation

• multimodal reasoning
• guideline-aligned diagnosis
• reinforcement fine-tuning
• cross-attention fusion
• explainable AI