Prominent researchers from NYU Grossman School of Medicine, along with collaborators at the University of Michigan, led a new study demonstrating how a novel diagnostic approach combining optimal imaging with an artificial intelligence (AI) algorithm can produce accurate, real-time interoperative diagnosis of brain tumors.
This prospective study was designed to compare the diagnostic accuracy of stimulated Raman histology (SRH) brain tumor image classification through machine learning, along with that of pathologist interpretation of conventional histologic images. The study was published in Nature Medicine in a paper titled ‘Near real-time interoperative brain tumor diagnosis using stimulated Raman histology and deep neural networks.’
About 80% of the 15.2 million people worldwide who are diagnosed with cancer will undergo surgery and often parts of the removed tumor will be analyzed during surgery—partly to help offer a preliminary diagnosis. In America, more than 1.1 million biopsy specimens a year are taken. The existing workflow for interoperative diagnosis based on hematoxylin and eosin staining of processed tissue is time, resource and labor-intensive. The conventional workflow for interoperative histology actually dates back over a century! It is a manual, labor-intensive, and laborious process.
The New Approach
Developed by the author of this study (Orringer) and his colleagues, the team exploits advances in optics and AI. The SRH imaging technique offers label-free, sub-micrometer-resolution images of unprocessed biologic tissues to reveal tumor infiltration.
The Results of the Study
The results indicated that the AI-based diagnosis was 94.6% accurate, with pathologist-based interpretation demonstrating 93.9% accuracy. The study authors articulate that the system’s precise diagnostic capacity could benefit those centers that are lacking in expert neuropathologists, reports Genetic Engineering and Biotechnology News.
Daniel A. Orringer, MD associate professor of neurosurgery, NYU Grossman School of Medicine