Differential dementia diagnosis
Transformer that uses multimodal clinical data to distinguish between dementia etiologies.
Open science
Tools & models
We release code, models, datasets and demos alongside our papers so others can reproduce, validate and build on our work. Code is on GitHub; models, datasets and interactive demos are on Hugging Face.
Featured
Alzheimer's biomarker assessment
Fuses routinely collected multimodal data to estimate amyloid and tau PET status.
Multimodal AD dementia assessment
Deep learning on MRI, demographics, history and neuropsychological tests to assess cognitive status.
Interpretable AD classification
Interpretable deep learning framework linking MRI to Alzheimer's disease with disease probability maps.
Graph-transformer for WSIs
Graph-transformer for whole slide image classification.
PodGPT
Audio-augmented large language model for research and education.
More code
Pathology + gene expression fusion
Graph attention-based fusion of pathology images and gene expression for cancer survival prediction.
FourierMIL
Fourier filtering-based multiple instance learning for whole slide image analysis.
Agentic medical QA
Agentic memory-augmented retrieval and evidence grounding for medical question answering.
Explainability-driven feature acquisition
Active feature acquisition that ranks what to measure next using explainability.
Disease-driven domain generalization
Domain generalization for neuroimaging-based assessment of Alzheimer's disease.
Dementia detection from voice
Deep learning on voice recordings from the Framingham Heart Study to detect dementia.
Speech features on GPU
A re-implementation of Python Speech Features with GPU acceleration.
Medical AI Hackathon
Teaching materials from our medical AI hackathon.
On Hugging Face
Interactive demos, model weights and datasets at huggingface.co/vkola-lab.
Demo
vkola-lab/nmed2024
Differential dementia diagnosis demo
Interactive demo of the Nature Medicine (2024) model.
Demo vkola-lab/NCOMMS2025Alzheimer's biomarker assessment demo
Interactive demo of the Nature Communications (2025) model.
Models vkola-lab/PodGPT-*PodGPT models
PodGPT fine-tuned from Llama 3.1/3.2, Gemma 2 and Mixtral, from 1B to 70B parameters.
Dataset vkola-lab/NLM-LitArch-…-TextbooksNLM LitArch open-access textbooks
8,226 open-access medical textbooks used for agentic medical retrieval.
Dataset vkola-lab/PubMed-Central-ArticlesPubMed Central articles
Creative Commons research articles from leading biomedical journals.
Dataset vkola-lab/ClinicalTrials.govClinicalTrials.gov records
Processed records of completed clinical trials.
Looking for something else? Browse everything at github.com/vkola-lab and huggingface.co/vkola-lab. For questions about using our models or collaborating, see Contact.