Faculty Bibliography

The AI era in medicine has ushered in new opportunities to improve the diagnosis and treatment of human disease. CHARM, an AI algorithm described in this issue,¹ has the potential to streamline molecular classification, intraoperative diagnosis, surgical decision making, and trial enrollment for glioma patients.

INTRODUCTION/BACKGROUND:Delay between targeted prostate biopsy (PB) and pathologic diagnosis can lead to a concern of inadequate sampling and repeated biopsy. Stimulated Raman histology (SRH) is a novel microscopic technique allowing real-time, label-free, high-resolution microscopic images of unprocessed, unsectioned tissue. This technology holds potential to decrease the time for PB diagnosis from days to minutes. We evaluated the concordance of pathologist interpretation of PB SRH as compared with traditional hematoxylin and eosin (H&E) stained slides. METHODS:, to create SRH images. The cores were then processed as per normal pathologic protocols. Sixteen PB containing a mix of benign and malignant histology were used as an SRH training cohort for four genitourinary pathologists, who were then tested on a set of 32 PBs imaged by SRH and processed by traditional H&E. Sensitivity, specificity, accuracy, and concordance for prostate cancer (PCa) detection on SRH relative to H&E were assessed. RESULTS:The mean pathologist accuracy for the identification of any PCa on PB SRH was 95.7%. In identifying any PCa or ISUP grade group 2-5 PCa, a pathologist was independently able to achieve good and very good concordance (κ: 0.769 and 0.845, respectively; p < 0.001). After individual assessment was completed a pathology consensus conference was held for the interpretation of the PB SRH; after the consensus conference the pathologists' concordance in identifying any PCa was also very good (κ: 0.925, p < 0.001; sensitivity 95.6%; specificity 100%). CONCLUSION/CONCLUSIONS:SRH produces high-quality microscopic images that allow for accurate identification of PCa in real-time without need for sectioning or tissue processing. The pathologist performance improved through progressive training, showing that ultimately high accuracy can be obtained. Ongoing SRH evaluation in the diagnostic and treatment setting hold promise to reduce time to tissue diagnosis, while interpretation by convolutional neural network may further improve diagnostic characteristics and broaden use.

INTRODUCTION/BACKGROUND:Renal tumor biopsy requires adequate tissue sampling to aid in the investigation of small renal masses. In some centers the contemporary nondiagnostic renal mass biopsy rate may be as high as 22% and may be as high as 42% in challenging cases. Stimulated Raman Histology (SRH) is a novel microscopic technique which has created the possibility for rapid, label-free, high-resolution images of unprocessed tissue which may be viewed on standard radiology viewing platforms. The application of SRH to renal biopsy may provide the benefits of routine pathologic evaluation during the procedure, thereby reducing nondiagnostic results. We conducted a pilot feasibility study, to assess if renal cell carcinoma (RCC) subtypes may be imaged and to see if high-quality hematoxylin and eosin (H&E) could subsequently be generated. METHODS/MATERIALS/METHODS:. The cores were then processed as per routine pathologic protocols. The SRH images and hematoxylin and eosin (H&E) slides were then viewed by a genitourinary pathologist. RESULTS:The SRH microscope took 8 to 11 minutes to produce high-quality images of the renal biopsies. Total of 25 renal tumors including 1 oncocytoma, 3 chromophobe RCC, 16 clear cells RCC, 4 papillary RCC, and 1 medullary RCC were included. All renal tumor subtypes were captured, and the SRH images were easily differentiated from adjacent normal renal parenchyma. High quality H&E slides were produced from each of the renal biopsies after SRH was completed. Immunostains were performed on selected cases and the staining was not affected by the SRH image process. CONCLUSION/CONCLUSIONS:SRH produces high quality images of all renal cell subtypes that can be rapidly produced and easily interpreted to determine renal mass biopsy adequacy, and on occasion, may allow renal tumor subtype identification. Renal biopsies remained available to produce high quality H&E slides and immunostains for confirmation of diagnosis. Procedural application has promise to decrease the known rate of renal mass nondiagnostic biopsies, and application of convolutional neural network methodology may further improve diagnostic capability and increase utilization of renal mass biopsy among urologists.

Learning high-quality, self-supervised, visual representations is essential to advance the role of computer vision in biomedical microscopy and clinical medicine. Previous work has focused on self-supervised representation learning (SSL) methods developed for instance discrimination and applied them directly to image patches, or fields-of-view, sampled from gigapixel whole-slide images (WSIs) used for cancer diagnosis. However, this strategy is limited because it (1) assumes patches from the same patient are independent, (2) neglects the patient-slide-patch hierarchy of clinical biomedical microscopy, and (3) requires strong data augmentations that can degrade downstream performance. Importantly, sampled patches from WSIs of a patient's tumor are a diverse set of image examples that capture the same underlying cancer diagnosis. This motivated HiDisc, a data-driven method that leverages the inherent patient-slide-patch hierarchy of clinical biomedical microscopy to define a hierarchical discriminative learning task that implicitly learns features of the underlying diagnosis. HiDisc uses a self-supervised contrastive learning framework in which positive patch pairs are defined based on a common ancestry in the data hierarchy, and a unified patch, slide, and patient discriminative learning objective is used for visual SSL. We benchmark HiDisc visual representations on two vision tasks using two biomedical microscopy datasets, and demonstrate that (1) HiDisc pretraining outperforms current state-of-the-art self-supervised pretraining methods for cancer diagnosis and genetic mutation prediction, and (2) HiDisc learns high-quality visual representations using natural patch diversity without strong data augmentations.

Stimulated Raman histology (SRH) is an ex vivo optical imaging method that enables microscopic examination of fresh tissue intraoperatively. The conventional intraoperative method uses frozen section analysis, which is labor and time intensive, introduces artifacts that limit diagnostic accuracy, and consumes tissue. SRH imaging allows rapid microscopic imaging of fresh tissue, avoids tissue loss, and enables remote telepathology review. This improves access to expert neuropathology consultation in both low- and high-resource practices. We clinically validated SRH by performing a blinded, retrospective two-arm telepathology study to clinically validate SRH for telepathology at our institution. Using surgical specimens from 47 subjects, we generated a data set composed of 47 SRH images and 47 matched whole slide images (WSIs) of formalin-fixed, paraffin-embedded tissue stained with hematoxylin and eosin, with associated intraoperative clinicoradiologic information and structured diagnostic questions. We compared diagnostic concordance between WSI and SRH-rendered diagnoses. Also, we compared the 1-year median turnaround time (TAT) of intraoperative conventional neuropathology frozen sections with prospectively rendered SRH-telepathology TAT. All SRH images were of sufficient quality for diagnostic review. A review of SRH images showed high accuracy in distinguishing glial from nonglial tumors (96.5% SRH vs 98% WSIs) and predicting final diagnosis (85.9% SRH vs 93.1% WSIs). SRH-based diagnosis and WSI-permanent section diagnosis had high concordance (κ = 0.76). The median TAT for prospectively SRH-rendered diagnosis was 3.7 minutes, approximately 10-fold shorter than the median frozen section TAT (31 minutes). The SRH-imaging procedure did not affect ancillary studies. SRH generates diagnostic virtual histologic images with accuracy comparable to conventional hematoxylin and eosin-based methods in a rapid manner. Our study represents the largest and most rigorous clinical validation of SRH to date. It supports the feasibility of implementing SRH as a rapid method for intraoperative diagnosis complementary to conventional pathology laboratory methods.

INTRODUCTION:Molecular classification has transformed the management of brain tumors by enabling more accurate prognostication and personalized treatment. Access to timely molecular diagnostic testing for brain tumor patients is limited, complicating surgical and adjuvant treatment and obstructing clinical trial enrollment. METHODS:By combining stimulated Raman histology (SRH), a rapid, label-free, non-consumptive, optical imaging method, and deep learning-based image classification, we are able to predict the molecular genetic features used by the World Health Organization (WHO) to define the adult-type diffuse glioma taxonomy, including IDH-1/2, 1p19q-codeletion, and ATRX loss. We developed a multimodal deep neural network training strategy that uses both SRH images and large-scale, public diffuse glioma genomic data (i.e. TCGA, CGGA, etc.) in order to achieve optimal molecular classification performance. RESULTS:One institution was used for model training (University of Michigan) and four institutions (NYU, UCSF, Medical University of Vienna, and University Hospital Cologne) were included for patient enrollment in the prospective testing cohort. Using our system, called DeepGlioma, we achieved an average molecular genetic classification accuracy of 93.2% and identified the correct diffuse glioma molecular subgroup with 91.5% accuracy within 2 minutes in the operating room. DeepGlioma outperformed conventional IDH1-R132H immunohistochemistry (94.2% versus 91.4% accuracy) as a first-line molecular diagnostic screening method for diffuse gliomas and can detect canonical and non-canonical IDH mutations. CONCLUSIONS:Our results demonstrate how artificial intelligence and optical histology can be used to provide a rapid and scalable alternative to wet lab methods for the molecular diagnosis of brain tumor patients during surgery.

Molecular classification has transformed the management of brain tumors by enabling more accurate prognostication and personalized treatment. However, timely molecular diagnostic testing for patients with brain tumors is limited, complicating surgical and adjuvant treatment and obstructing clinical trial enrollment. In this study, we developed DeepGlioma, a rapid (<90 seconds), artificial-intelligence-based diagnostic screening system to streamline the molecular diagnosis of diffuse gliomas. DeepGlioma is trained using a multimodal dataset that includes stimulated Raman histology (SRH); a rapid, label-free, non-consumptive, optical imaging method; and large-scale, public genomic data. In a prospective, multicenter, international testing cohort of patients with diffuse glioma (n = 153) who underwent real-time SRH imaging, we demonstrate that DeepGlioma can predict the molecular alterations used by the World Health Organization to define the adult-type diffuse glioma taxonomy (IDH mutation, 1p19q co-deletion and ATRX mutation), achieving a mean molecular classification accuracy of 93.3 ± 1.6%. Our results represent how artificial intelligence and optical histology can be used to provide a rapid and scalable adjunct to wet lab methods for the molecular screening of patients with diffuse glioma.

BACKGROUND/UNASSIGNED:Central nervous system (CNS) cancer is the 10th leading cause of cancer-associated deaths for adults, but the leading cause in pediatric patients and young adults. The variety and complexity of histologic subtypes can lead to diagnostic errors. DNA methylation is an epigenetic modification that provides a tumor type-specific signature that can be used for diagnosis. METHODS/UNASSIGNED:We performed a prospective study using DNA methylation analysis as a primary diagnostic method for 1921 brain tumors. All tumors received a pathology diagnosis and profiling by whole genome DNA methylation, followed by next-generation DNA and RNA sequencing. Results were stratified by concordance between DNA methylation and histopathology, establishing diagnostic utility. RESULTS/UNASSIGNED:Of the 1602 cases with a World Health Organization histologic diagnosis, DNA methylation identified a diagnostic mismatch in 225 cases (14%), 78 cases (5%) did not classify with any class, and in an additional 110 (7%) cases DNA methylation confirmed the diagnosis and provided prognostic information. Of 319 cases carrying 195 different descriptive histologic diagnoses, DNA methylation provided a definitive diagnosis in 273 (86%) cases, separated them into 55 methylation classes, and changed the grading in 58 (18%) cases. CONCLUSIONS/UNASSIGNED:DNA methylation analysis is a robust method to diagnose primary CNS tumors, improving diagnostic accuracy, decreasing diagnostic errors and inconclusive diagnoses, and providing prognostic subclassification. This study provides a framework for inclusion of DNA methylation profiling as a primary molecular diagnostic test into professional guidelines for CNS tumors. The benefits include increased diagnostic accuracy, improved patient management, and refinements in clinical trial design.

OBJECTIVE:Intraoperative neuropathological assessment with conventional frozen sections supports the neurosurgeon in optimizing the surgical strategy. However, preparation and review of frozen sections can take as long as 45 minutes. Stimulated Raman histology (SRH) was introduced as a novel technique to provide rapid high-resolution digital images of unprocessed tissue samples directly in the operating room that are comparable to conventional histopathological images. Additionally, SRH images are simultaneously and easily accessible for neuropathological judgment. Recently, the first study showed promising results regarding the accuracy and feasibility of SRH compared with conventional histopathology. Thus, the aim of this study was to compare SRH with conventional H&E images and frozen sections in a large cohort of patients with different suspected central nervous system (CNS) tumors. METHODS:The authors included patients who underwent resection or stereotactic biopsy of suspected CNS neoplasm, including brain and spinal tumors. Intraoperatively, tissue samples were safely collected and SRH analysis was performed directly in the operating room. To enable optimal comparison of SRH with H&E images and frozen sections, the authors created a digital databank that included images obtained with all 3 imaging modalities. Subsequently, 2 neuropathologists investigated the diagnostic accuracy, tumor cellularity, and presence of diagnostic histopathological characteristics (score 0 [not present] through 3 [excellent]) determined with SRH images and compared these data to those of H&E images and frozen sections, if available. RESULTS:In total, 94 patients with various suspected CNS tumors were included, and the application of SRH directly in the operating room was feasible in all cases. The diagnostic accuracy based on SRH images was 99% when compared with the final histopathological diagnosis based on H&E images. Additionally, the same histopathological diagnosis was established in all SRH images (100%) when compared with that of the corresponding frozen sections. Moreover, the authors found a statistically significant correlation in tumor cellularity between SRH images and corresponding H&E images (p < 0.0005 and R = 0.867, Pearson correlation coefficient). Finally, excellent (score 3) or good (2) accordance between diagnostic histopathological characteristics and H&E images was present in 95% of cases. CONCLUSIONS:The results of this retrospective analysis demonstrate the near-perfect diagnostic accuracy and capability of visualizing relevant histopathological characteristics with SRH compared with conventional H&E staining and frozen sections. Therefore, digital SRH histopathology seems especially useful for rapid intraoperative investigation to confirm the presence of diagnostic tumor tissue and the precise tumor entity, as well as to rapidly analyze multiple tissue biopsies from the suspected tumor margin. A real-time analysis comparing SRH images and conventional histological images at the time of surgery should be performed as the next step in future studies.

Accurate intraoperative diagnosis is essential for providing safe and effective care during brain tumor surgery. Our standard-of-care diagnostic methods are time, resource, and labor intensive, which restricts access to optimal surgical treatments. To address these limitations, we propose an alternative workflow that combines stimulated Raman histology (SRH), a rapid optical imaging method, with deep learning-based automated interpretation of SRH images for intraoperative brain tumor diagnosis and real-time surgical decision support. Here, we present OpenSRH, the first public dataset of clinical SRH images from 300+ brain tumors patients and 1300+ unique whole slide optical images. OpenSRH contains data from the most common brain tumors diagnoses, full pathologic annotations, whole slide tumor segmentations, raw and processed optical imaging data for end-to-end model development and validation. We provide a framework for patch-based whole slide SRH classification and inference using weak (i.e. patient-level) diagnostic labels. Finally, we benchmark two computer vision tasks: multiclass histologic brain tumor classification and patch-based contrastive representation learning. We hope OpenSRH will facilitate the clinical translation of rapid optical imaging and real-time ML-based surgical decision support in order to improve the access, safety, and efficacy of cancer surgery in the era of precision medicine. Dataset access, code, and benchmarks are available at https://opensrh.mlins.org.