Intraoperative label-free tissue diagnostics using a stimulated Raman histology imaging system with artificial intelligence: An initial experience.

Abstract

Background: Accurate intraoperative tissue diagnostics could impact on decision making regarding the extent of resection (EOR) during brain tumor surgery. Stimulated Raman histology (SRH) is a label-free optical imaging method that uses different biochemical properties of tissue to generate a hematoxylin-eosin-like image and, in combination with an artificial intelligence-based image classifier, offers the opportunity to obtain rapid intraoperative tissue diagnoses.

Objective: The goal of this study was to report on our initial experience with SRH to evaluate its accuracy in comparison to final tissue diagnosis.

Materials & methods: We evaluated 70 consecutive adult cases with brain tumors. We compared results of the three different SRH classifier (diagnostic, molecular and tumor/non-tumor) to the respective final histopathological result. Similarly, we evaluated the isocitrate dehydrogenase (IDH) mutations in 18 patients using SRH. Lastly, we compared SRH results of samples taken from the tumor margins with early postoperative MRI. Prediction accuracy was evaluated by logistic regression and Receiver Operator Curve (ROC) analysis.

Results: We included 19 gliomas, 9 metastases, 22 meningiomas and 14 other tumor entities. Regarding accuracy of intraoperative SRH predictions, regression analysis showed an Area Under the Curve (AUC) of 0.77 (95 % C.I. 0.64-0.89, p = 0.0008), suggesting agreement of predictions with final diagnosis. For specific tumor entities, variable accuracies were observed: The highest accuracy was obtained for meningiomas followed by high-grade glioma. IDH mutations were predicted with an AUC of 0.93 (95 % C.I. 0.88-0.98; p < 0.0001). The SRH examination of tissue samples from tumor margins corresponded with postoperative MRI in 4 out of 5 cases.

Conclusion: Our initial experience with SRH shows that this novel imaging technique is a promising approach to obtain rapid intraoperative tissue diagnosis to guide surgical decision making based on histology and cell-density. With further refinement of AI-based automated image classification and a better integration into the surgical workflow, prediction accuracy and reliability could be improved.