NeuroVFM: Health system learning enables generalist neuroimaging models

Paper / Interactive Demo / GitHub / MLiNS Lab

This is the model card for the NeuroVFM MRI diagnosis model, a modified AB-MIL model that diagnoses an expert-defined clinical ontology of 74 diagnoses, trained on 218,882 MRI neuroimaging studies from more than 20 years of routine clinical care at a large academic center.

The feature backbone, NeuroVFM, can be found here.

Model Details

• Architecture: Modified AB-MIL ('classify-then-aggregate')
• Training Data: UM-NeuroImages MRI
  • Diversity: 218,882 unique MRI studies
• Training Objective: Multilabel binary classification
• Compute Hardware: Trained on 8x NVIDIA L40S GPUs (48GB VRAM)
• Training Efficiency: <100 GPU-hours total pretraining time (Automatic Mixed Precision with PyTorch DDP)
• Optimization: AdamW, LR of 5e-4 with Cosine Decay (10% warmup)

Quick Start

The easiest way to use this diagnostic model is through our Python package.

Limitations & Safety

This model is a research tool. It has not been approved by the FDA or any regulatory body for clinical use. While trained on a diverse health system population, the model may carry biases intrinsic to the University of Michigan patient cohort. When used for generation (with an LLM), the system may still hallucinate findings, though at a lower rate than pure language models. Outputs must be verified by a clinician.

License

• Weights: CC-BY-NC-SA 4.0 (Non-Commercial Research Use)
• Code: MIT License