AI reads MRIs to predict brain tumor genetics — no biopsy needed

What if AI could read routine MRI scans and predict a brain tumor’s genetics — no biopsy? This study pairs computational imaging with large language models (LLMs) to zero-shot predict IDH mutation status in gliomas.

• Multi-parametric MRI plus tumor segmentation become interpretable visual attributes and quantitative features.
• Those features are serialized into a standardized JSON and sent as a prompt — no model fine-tuning.
• Tested on six public datasets (N=1,427): strong, balanced performance across varied cohorts, even without manual annotations.
• GPT-5 outperformed GPT-4o for context-driven phenotype interpretation.
• Volumetric features mattered most; subtype-specific imaging markers and clinical info added value.

Why it matters: toward precise, non-invasive tumor genotyping and faster, safer decisions in neuro-oncology. Preprint: http://arxiv.org/abs/2511.03376v1 Code: https://github.com/ATPLab-LUMS/CIM-LLM

Paper: http://arxiv.org/abs/2511.03376v1

Register: https://www.AiFeta.com

AI MedicalImaging NeuroOncology MRI Radiogenomics LLM BrainTumor IDH Healthcare OpenScience