Rapid Automated Analysis of Skull Base Tumor Specimens Using Intraoperative Optical Imaging and Artificial Intelligence

Abstract

BACKGROUND:Accurate specimen analysis of skull base tumors is essential for providing personalized surgical treatment strategies. Intraoperative specimen interpretation can be challenging because of the wide range of skull base pathologies and lack of intraoperative pathology resources.OBJECTIVE:To develop an independent and parallel intraoperative workflow that can provide rapid and accurate skull base tumor specimen analysis using label-free optical imaging and artificial intelligence.METHODS:We used a fiber laser–based, label-free, nonconsumptive, high-resolution microscopy method (<60 seconds per 1 × 1 mm2), called stimulated Raman histology (SRH), to image a consecutive, multicenter cohort of patients with skull base tumor. SRH images were then used to train a convolutional neural network model using 3 representation learning strategies: cross-entropy, self-supervised contrastive learning, and supervised contrastive learning. Our trained convolutional neural network models were tested on a held-out, multicenter SRH data set.RESULTS:SRH was able to image the diagnostic features of both benign and malignant skull base tumors. Of the 3 representation learning strategies, supervised contrastive learning most effectively learned the distinctive and diagnostic SRH image features for each of the skull base tumor types. In our multicenter testing set, cross-entropy achieved an overall diagnostic accuracy of 91.5%, self-supervised contrastive learning 83.9%, and supervised contrastive learning 96.6%. Our trained model was able to segment tumor-normal margins and detect regions of microscopic tumor infiltration in meningioma SRH images.CONCLUSION:SRH with trained artificial intelligence models can provide rapid and accurate intraoperative analysis of skull base tumor specimens to inform surgical decision-making.

Novel rapid intraoperative qualitative tumor detection by a residual convolutional neural network using label-free stimulated Raman scattering microscopy

“…It was therefore chosen for further validation and analysis. Other studies showed similar overall accuracy as well as ROC-AUC ranging from approximately 84 to 97% between the CNN- and neuropathological-based SRH image analysis [ 4 , 5 , 15 , 16 ]. However, it should be noted that the CNN deployed in these studies had different output classes with different loss functions.…”

OpenSRH: optimizing brain tumor surgery using intraoperative stimulated Raman histology

“…One application of SRH is the detection of tumor infiltration in real-time to improve the extent of tumor resection and reduce residual tumor burden. Real-time SRH-based tumor delineation has been studied in sinonasal/skull base cancers [13,29,30] and diffuse gliomas [19,31]. OpenSRH provides the necessary dataset to explore this topic for multiple brain tumor types, including metastatic tumors and extra-axial tumors, such as meningiomas (Figure 1).…”

“…Standardized: SRH image acquisition is invariant to patient demographic features, clinical workforce, and geographic location 4. Accurate: preliminary results [6,13] and diagnostic performance benchmarks (see Figure 4) are on par with the pathologist-based interpretation of H&E histology…”

“…Previous studies showed that SRH plus shallow ML models can be used to detect and quantify tumor infiltration in adult and pediatric fresh surgical specimens [5,18,19,20]. We subsequently demonstrated that SRH combined with convolutional neural network architectures can be used for intraoperative diagnostic decision support [6,12,13]. These preliminary studies, while demonstrating the feasibility of applying deep architectures to SRH, did not include rigorous hyperparameter tuning, explicit representation learning, or ablation studies to optimize model performance.…”

“…We propose an innovative diagnosis system that uses a rapid (2-3 minutes, >10× speedup), label-free optical histology method, called stimulated Raman histology (SRH), combined with deep learning-based interpretation of fresh, unprocessed surgical specimens. We have previously demonstrated the feasibility of large-scale clinical SRH imaging [5] and the use of deep neural networks for SRH image classification of brain tumor patients [6,12,13]. These studies demonstrate the potential for AI-based diagnosis and interpretation of SRH images to better inform brain tumor surgery and provide personalized surgical goals in the era of precision medicine.…”