It's time to rethink how we help patients

Remembrance is an emotionally intelligent, context-aware LLM-based memory assistant for Alzheimer's patients. It features a user interface similar to Obsidian and delivers reminiscence therapy in a non-intrusive manner. The system preserves core memories by extracting data from conversations using advanced language models.

Clarity is a deep learning framework that enhances low-field MRI scans for segmenting brain regions linked to Alzheimer's Disease. It achieved 96% accuracy in automatic AD diagnosis using volumetric data and ensemble ML models. This approach offers a cost-effective, faster alternative to traditional MRI for early AD detection.

GeneAttentionNet classifies Alzheimer's disease using high-dimensional gene expression data with attention mechanisms to uncover meaningful gene relationships. Unlike traditional models treating genes as isolated features, it mimics biological system interactions. The model outperformed standard multilayer perceptrons by nearly 8% and Random Forests by 38%.

SALSA is a lightweight framework for detecting Alzheimer's from spontaneous speech while maintaining strong generalization across datasets. It uses frozen pre-trained encoders with LoRA adapters and combines acoustic embeddings with lexical cues. Designed to run efficiently on consumer hardware, it demonstrates strong cross-corpus performance on benchmarks like ADReSS 2020.

NeuroMorphLite-fMRI merges deep learning and reinforcement learning to enhance fMRI-based Alzheimer's research with low computational demands. Built on ADNI and OASIS datasets, it uses CNN and Transformer backbones for high-fidelity brain reconstruction. A reinforcement agent optimizes regularization and identifies key regions like the hippocampus for both image clarity and diagnostic accuracy.