For accurate, real-time intraoperative diagnosis of brain tumors, researchers have merged an advanced optical imaging technique with an AI system.
The accuracy of pathologist interpretation of traditional histologic images was compared to the diagnostic accuracy of brain tumor image categorization through machine learning. Both methods produced comparable findings, with the AI-based diagnosis being 94.6% accurate and the pathologist-based interpretation being 93.9% accurate.
By gathering scattered laser light, the imaging method known as stimulated Raman histology (SRH) shows tumor infiltration in human tissue while highlighting crucial details typically hidden in conventional histologic images. After processing and analyzing the microscopic images using AI, surgeons can view a predicted brain tumor diagnostic in under 2.5 minutes. After the resection, they can precisely find and remove a tumor that would have been invisible using the same technology.
The researchers used more than 2.5 million samples from 415 patients to train a deep convolutional neural network (CNN) to classify tissue into 13 histologic categories representing the most prevalent brain tumors, including malignant gliomas, lymphomas, metastatic tumors, and meningiomas.
Researchers recruited 278 patients having brain tumor removal or epilepsy surgery at three university medical centers for the prospective clinical trial to validate the CNN.
Related Content: Terahertz Imaging Spots Microscopic Twists In Tissues