From RuntimeWire · RuntimeWire · · 40s
USC Researchers Release Alzheimer's Model for Missing Scans
USC researchers publish Alzheimer's model that works with incomplete brain scans.
In brief
USC researchers publish Alzheimer's model that works with incomplete brain scans. USC researchers created MEMOIR-VLM to classify Alzheimer's using incomplete scans, achieving 91.3% accuracy on research data. Originally reported by RuntimeWire.
USC Researchers Publish Alzheimer's Model
USC researchers created MEMOIR-VLM to classify Alzheimer's using incomplete brain scans and clinical data.
Model Achieves 91.3% Accuracy
The model distinguishes cognitively normal people from dementia patients with 91.3% balanced accuracy on held-out research data.
Limitations of the Model
Cognitive scores drive most diagnostic performance, and the language model did not beat nearest-neighbor retrieval.
What matters most
USC researchers created MEMOIR-VLM to classify Alzheimer's using incomplete scans, achieving 91.3% accuracy on research data.
