SpineXtract, an AI-driven system using stimulated Raman histology (SRH), achieved a 92.9% balanced accuracy in classifying common spinal tumors intraoperatively within 5 minutes. This multicenter study demonstrated consistent performance across institutions, outperforming existing brain tumor classifiers by over 15%.
Background
Intraoperative diagnosis of spinal tumors is critical for guiding surgical and treatment decisions but is currently limited by time-consuming frozen section analysis requiring specialized pathology expertise. MRI often cannot definitively classify spinal tumors intraoperatively, and histopathology results take days. Stimulated Raman Histology (SRH) offers rapid, label-free imaging without tissue processing, and AI integration can facilitate near-real-time diagnosis. Prior AI classifiers exist for brain tumors but none specialized for spinal tumors until SpineXtract.
Data Highlights
Metric
Value
95% Confidence Interval
Macro-average balanced accuracy
92.9%
85.5–98.2%
Tumor-specific accuracy range
84.2–98.6%
Institutional accuracy range
91.4–92.0%
Number of patients
44
Number of SRH slide images
142
Key Findings
SpineXtract achieved 92.9% macro-average balanced accuracy in classifying meningioma, schwannoma, ependymoma, and metastasis intraoperatively.
Diagnostic performance was consistent across three international tertiary centers (NYU, UM, MUV) with balanced accuracy between 91.4% and 92.0%.
The AI system provided results within 5 minutes, significantly faster than traditional frozen section pathology.
SpineXtract outperformed existing brain tumor AI classifiers by 15.6% in spinal tumor classification accuracy.
The system generated interpretable visual heatmaps to assist surgeons and pathologists during surgery.
No adverse events occurred related to tissue sampling for SRH imaging, confirming safety and feasibility.
Clinical Implications
SpineXtract offers a rapid, accurate, and label-free intraoperative diagnostic tool that can reduce reliance on time-intensive frozen section pathology and specialized pathologists. Its consistent performance across institutions supports broad clinical applicability, potentially improving surgical decision-making and patient outcomes in spinal tumor management. The system's visual feedback enhances intraoperative interpretation, facilitating real-time treatment adjustments.
Conclusion
The development and validation of SpineXtract demonstrate that AI-powered label-free Raman spectromics can transform intraoperative spinal tumor diagnosis by providing rapid, accurate, and interpretable results. This technology holds promise to streamline surgical workflows and improve diagnostic precision in spinal oncology.
References
Article Source 2024 -- AI-Powered Label-Free Raman Spectromics for Real-Time Assessment of Spinal Tumors During Surgery
by David Reinecke, Nina Müller, Anna-Katharina Meissner, Gina Fürtjes, Lili Leyer, Claire Wang, Adrian Ion-Margineanu, Nader Maarouf, Andrew Smith, Todd C. Hollon, Cheng Jiang, Xinhai Hou, Abdulkader Al-Shughri, Lisa I. Körner, Georg Widhalm, Thomas Roetzer-Pejrimovsky, Matija Snuderl, Sandra Camelo-Piragua, John G. Golfinos, Roland Goldbrunner, Daniel A. Orringer, Niklas von Spreckelsen, Volker Neuschmelting
