arXiv:2605.28397v1 Announce Type: new 
Abstract: Predicting conversion from Mild Cognitive Impairment (MCI) to Alzheimer's Disease (AD) is critical for early intervention. Current deep learning paradigms predominantly rely on cross-sectional structural MRI, neglecting prognostic value in patient-specific anatomical trajectories. We introduce the Temporal Adaptive Fusion Network (TAF-Net), a hybrid CNN-Transformer architecture that models paired longitudinal 3D MRI scans. Central to TAF-Net is a Temporal Fusion Module governed by an Adaptive Temporal Gate, which learns patient-specific weightings to synthesize three spatiotemporal representations: explicit structural change, region-to-region temporal cross-attention, and bilateral feature concatenation. Evaluated on the Alzheimer's Disease Neuroimaging Initiative cohort for three-year MCI-to-AD conversion prediction, TAF-Net achieved the highest discriminative performance among all evaluated methods using only structural MRI, significantly outperforming the strongest baseline and approaching multimodal methods requiring PET, CSF, or genetic data. The architecture exhibited exceptional data efficiency, matching baseline performance with a fraction of training data. Ablation studies demonstrate that longitudinal fusion improves discrimination while reducing predictive variance by 48% compared to single-timepoint evaluation. Interpretability analyses reveal spatial attention aligned with established AD pathology in the medial temporal lobe and ventricles, while the gating mechanism prioritizes explicit volumetric change with strong positive correlation to conversion risk.

تقدم دراسة جديدة شبكة الدمج الزمني التكيفية (TAF-Net)، وهي بنية هجينة من CNN-Transformer مصممة للتنبؤ بالتحول من ضعف الإدراك المعتدل (MCI) إلى مرض الزهايمر (AD) باستخدام مسح MRI الطولي المزدوج. يركز هذا النهج المبتكر على المسارات التشريحية الخاصة بالمرضى، والتي تم تجاهلها إلى حد كبير في النماذج الحالية للتعلم العميق.

Un nuevo estudio presenta el Temporal Adaptive Fusion Network (TAF-Net), una arquitectura híbrida de CNN-Transformer diseñada para predecir la conversión de Deterioro Cognitivo Leve (MCI) a la Enfermedad de Alzheimer (AD) utilizando escaneos de MRI longitudinales emparejados. Este enfoque innovador enfatiza las trayectorias anatómicas específicas del paciente, que han sido en gran medida pasadas por alto en los modelos actuales de aprendizaje profundo.

Une nouvelle étude présente le Temporal Adaptive Fusion Network (TAF-Net), une architecture hybride CNN-Transformer conçue pour prédire la conversion de l'Impairment Cognitif Léger (MCI) à la Maladie d'Alzheimer (AD) en utilisant des IRM longitudinales appariées. Cette approche innovante met l'accent sur les trajectoires anatomiques spécifiques aux patients, souvent négligées dans les modèles d'apprentissage profond actuels.

A new study introduces the Temporal Adaptive Fusion Network (TAF-Net), a hybrid CNN-Transformer architecture designed to predict the conversion from Mild Cognitive Impairment (MCI) to Alzheimer's Disease (AD) using paired longitudinal MRI scans. This innovative approach emphasizes patient-specific anatomical trajectories, which have been largely overlooked in current deep learning models.