Skip to main content
New: Get an independent second opinion on your scan Try SecondRead
Medical Imaging 2025-03-26

AI-Driven MRI Spine Pathology Detection

A comprehensive deep learning framework for automated MRI spine pathology detection across diverse clinical settings, using advanced neural network architectures and attention mechanisms to identify subtle spinal abnormalities with high sensitivity and specificity.

5C Network Research Team · arXiv · DOI: 10.48550/arXiv.2503.20316

Key findings

  • Trained on 2 million MRI spine scans from healthcare facilities across India, balanced by age, gender and scanner manufacturer
  • Classifies, segments and detects 43 distinct spinal pathologies
  • Up to 97.9% multi-pathology detection and 98.0% accuracy for normal vs abnormal classification
  • Deployed across 13 major Indian healthcare enterprises, processing approximately 100,000+ MRI spine scans

Plain-English summary

The question
Can an autonomous AI system detect the full range of common spinal pathologies on MRI, and keep performing consistently across patient groups and scanner brands?
The data
2 million MRI spine scans from diverse healthcare facilities across India, balanced across age groups, genders and scanner manufacturers.
The method
Vision Transformers, U-Net with cross-attention, MedSAM and Cascade R-CNN, together covering classification, segmentation and detection of 43 distinct spinal pathologies. Subgroup analyses checked performance by demographics, imaging conditions and equipment type.
The result
Up to 97.9% multi-pathology detection and 98.0% normal vs abnormal accuracy, consistent across age, gender and manufacturer subgroups. Deployed across 13 healthcare enterprises — diagnostic centers, large hospitals and government facilities — it processed approximately 100,000+ MRI spine scans, and the authors report reduced reporting times.

Why it matters for hospitals and radiologists

Evidence that one system holds up across scanner manufacturers matters for hospitals that run machines from more than one vendor. The deployment across 13 healthcare enterprises shows it working outside the lab, with the radiologist still reviewing every study.

Related at 5C: Bionic, the AI co-radiologist

Citation

Bargava Subramanian, Naveen Kumarasami, Praveen Shastry, Raghotham Sripadraj, Kalyan Sivasailam, Anandakumar D, Abinaya Ramachandran, Sudhir MP, Gunakutti G, Kishore Prasath Venkatesh. AI-Driven MRI Spine Pathology Detection: A Comprehensive Deep Learning Approach for Automated Diagnosis in Diverse Clinical Settings. arXiv preprint arXiv:2503.20316, 2025. https://doi.org/10.48550/arXiv.2503.20316

Read full paper on arXiv

Related research

From the lab to the worklist

This research runs in production.

The methods described here are part of Bionic, the AI layer that pre-reads every study 5C reports — 20,000+ scans a day across 2,000+ hospitals, each one signed by an NMC-registered radiologist.