The field of Alzheimer's disease research is moving towards the development of more accurate and interpretable diagnostic and prognostic models. Recent studies have focused on improving the accuracy of imaging biomarkers, such as those derived from MRI scans, and incorporating longitudinal clinical data to predict disease progression. The use of multi-task learning frameworks, vision transformers, and tabular foundation models has shown promise in predicting ADAS-Cog scores and other clinically relevant biomarkers. Additionally, there is a growing interest in developing explainable models that can uncover the complex relationships between brain structure, genetic variation, and disease progression. Noteworthy papers include:

• A study that developed an isotropic segmentation model for medial temporal lobe subregions, which showed higher significance in distinguishing between participants with mild cognitive impairment and cognitively unimpaired participants.
• A paper that proposed a weighted Vision Transformer-based multi-task learning framework for predicting ADAS-Cog scores, which demonstrated the importance of sub-score-specific loss weighting in improving model performance.
• A study that introduced NeuroPathX, an explainable deep learning framework for capturing meaningful interactions between structural variations in the brain and genetic variation.
• A paper that proposed EffNetViTLoRA, a hybrid deep learning approach for Alzheimer's disease diagnosis, which achieved high classification accuracy and F1-score across three diagnostic categories.