Modality Contribution Score - A Per-Patient Framework for Quantifying the Relative Diagnostic Contribution of Structural MRI and Amyloid PET in Alzheimer's Disease

AI can now explain which medical imaging type matters most for each patient: Rather than just giving a diagnosis, this new system (MCNet) shows whether a patient's Alzheimer's diagnosis is driven more by brain shrinkage visible on MRI or by amyloid buildup visible on PET scans—information that could help doctors personalize treatment decisions.

The balance between imaging types shifts predictably as Alzheimer's progresses: The research found a clear pattern where healthy people rely more on MRI findings (41% PET contribution), while Alzheimer's patients rely more on PET findings (67% PET contribution), suggesting the disease mechanism changes in measurable ways that AI can now track.

This framework provides actionable insights that blood tests alone cannot: While newer blood biomarkers for Alzheimer's are easier to obtain, this per-patient scoring system reveals which specific imaging abnormality is driving each individual's diagnosis—information that could guide whether to focus treatment on reducing amyloid versus addressing brain atrophy.

The system was validated across two large independent patient groups: The researchers tested their approach on 327 patients from one dataset and confirmed it worked similarly in 1,073 different patients from another dataset, suggesting the findings are reliable and could be used in real clinical settings.

This represents progress toward more transparent AI in medical imaging: By quantifying the contribution of each imaging modality per patient, this work addresses a key challenge in responsible AI—moving beyond "black box" diagnoses toward explainable decisions that clinicians can understand and trust.