Bhargava, Rohit

Abstract Histopathology has been a cornerstone of biomedical tissue assessment for decades, involving a labor-intensive and multi-step process that includes biopsy, gross examination, sampling, and subsequent tissue processing into snap-frozen or formalin-fixed paraffin-embedded (FFPE) blocks. These blocks are then sectioned, stained, and examined microscopically to diagnose and grade tissues. Despite its effectiveness, the traditional workflow is time-consuming, resource-intensive, and reliant on chemical dyes and expert interpretation. Recently, innovative chemical imaging techniques, such as Stimulated Raman Scattering (SRS) microscopy, have begun to redefine this field by enabling direct measurement of the molecular composition of tissues, thereby bypassing conventional tissue processing and staining protocols. This study demonstrates the integration of SRS microscopy into pathology workflows to rapidly acquire high-resolution chemical data from minimally processed fresh-frozen tissues. Unlike traditional methods that rely on thin-sectioned slides, this approach collects data from intact thick tissues using "optical sectioning, " enabling the generation of three-dimensional datasets across multiple planes. These chemical images are processed through a generative adversarial network (GAN) pipeline to produce virtual hematoxylin and eosin (H&E) images that are visually and diagnostically comparable to traditionally stained specimens. Furthermore, the generative model is extended to simulate the morphology and staining patterns of FFPE-processed tissues, offering a versatile tool for creating archival-quality images. This method significantly enhances diagnostic capabilities by preserving lipids and small molecules that are often lost during conventional processing. Within minutes, it generates high-quality stained images that demonstrated high inter-pathologist agreement in tissue grading, proving diagnostically reliable for clinical use. We evaluate our approach through a comprehensive survey of pathologists, assessing its diagnostic utility and reliability. Our results demonstrate that the generated images are comparable in quality to real histological images, showing no discernible inferiority in diagnostic utility. In prostate pathology, the integration of SRS microscopy and AI significantly streamlines traditional workflows, reducing complexity and processing time while preserving diagnostic precision. Citation Format: Kianoush Falahkheirkhah, Sudipta Mukherjee, John Cheville, Rohit Bhargava. Enhancing pathology workflows with chemical imaging and AI-driven generative models [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2469.

Pathology remains a labor-intensive discipline, relying on workflows rooted in practices established over a century ago. Despite incremental technological advancements, the diagnostic process—from tissue preparation to interpretation—still heavily depends on thinly sliced tissue sections stained and examined under brightfield microscopy by skilled pathologists. Difficult cases often necessitate iterative staining or complex immunohistochemistry (IHC) analyses. However, the recent emergence of machine learning (ML) presents an opportunity to revolutionize traditional diagnostic workflows. One promising advancement is Digital Chemical Pathology (DCP), a paradigm that combines modern ML algorithms with infrared (IR) imaging to streamline pathology and enhance diagnostic accuracy. Infrared wavelengths inherently align with molecular vibrational frequencies, providing intrinsic contrast based on the chemical composition of samples. Mid-IR spectroscopic microscopes leverage this principle to spatially resolve molecular distributions at the pixel level, creating "chemical fingerprints" that ML algorithms can correlate with cellular identities and disease states. Recent advancements, particularly the development of broadly tunable quantum cascade lasers (QCLs), have enabled faster IR imaging, paving the way for clinical applications of IR spectroscopic microscopes. In this research, we demonstrate the potential of DCP using frozen tissue samples. Our approach focuses on two core tasks: (1) virtual staining, translating IR images into realistic, stained images interpretable by pathologists, and (2) tissue segmentation, enabling the automated detection of cancer and the distinction between normal and cancerous tissues. Furthermore, we incorporate statistical safety algorithms to generate uncertainty maps, ensuring the reliability of predictions. Our findings highlight that the integration of IR imaging and ML can facilitate automated diagnostics, providing pathologists with powerful tools for enhanced healthcare delivery. By leveraging DCP, we aim to modernize pathology workflows, reduce diagnostic turnaround times, and improve patient outcomes through precise and efficient analysis. Citation Format: Kianoush Falahkheirkhah, Kevin Yeh, Ji-Hun Oh, Yen-Ting Liu, Ruo-Jing Ho, Ishaan Sharma, Matthew Confer, Andres Orr, Mykolas Tamonis, John Cheville, Rohit Bhargava. Transforming cancer diagnosis with digital chemical pathology [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2470.

ATM is a key mediator of radiation response, and pharmacological inhibition of ATM is a rational strategy to radiosensitize tumors. AZD1390 is a brain-penetrant ATM inhibitor and a potent radiosensitizer. This study evaluated the spectrum of radiosensitizing effects and the impact of TP53 mutation status in a panel of IDH1 wild-type (WT) glioblastoma (GBM) patient-derived xenografts (PDXs). AZD1390 suppressed radiation-induced ATM signaling, abrogated G 0 -G 1 arrest, and promoted a proapoptotic response specifically in p53-mutant GBM in vitro. In a preclinical trial using 10 orthotopic GBM models, AZD1390/RT afforded benefit in a cohort of TP53 -mutant tumors but not in TP53 -WT PDXs. In mechanistic studies, increased endogenous DNA damage and constitutive ATM signaling were observed in TP53 -mutant, but not in TP53 -WT, PDXs. In plasmid-based reporter assays, GBM43 ( TP53 -mutant) showed elevated DNA repair capacity compared with that in GBM14 (p53-WT), whereas treatment with AZD1390 specifically suppressed homologous recombination (HR) efficiency, in part, by stalling RAD51 unloading. Furthermore, overexpression of a dominant-negative TP53 (p53DD) construct resulted in enhanced basal ATM signaling, HR activity, and AZD1390-mediated radiosensitization in GBM14. Analyzing RNA-seq data from TCGA showed up-regulation of HR pathway genes in TP53 -mutant human GBM. Together, our results imply that increased basal ATM signaling and enhanced dependence on HR represent a unique susceptibility of TP53 -mutant cells to ATM inhibitor–mediated radiosensitization.

Histopathology has remained a cornerstone for biomedical tissue assessment for over a century, with a resource-intensive workflow involving biopsy or excision, gross examination, sampling, tissue processing to snap frozen or formalin-fixed paraffin-embedded blocks, sectioning, staining, optical imaging, and microscopic assessment. Emerging chemical imaging approaches, including stimulated Raman scattering (SRS) microscopy, can directly measure inherent molecular composition in tissue (thereby dispensing with the need for tissue processing, sectioning, and using dyes) and can use artificial intelligence (AI) algorithms to provide high-quality images. Here we show the integration of SRS microscopy in a pathology workflow to rapidly record chemical information from minimally processed fresh-frozen prostate tissue. Instead of using thin sections, we record data from intact thick tissues and use optical sectioning to generate images from multiple planes. We use a deep learning–based processing pipeline to generate virtual hematoxylin and eosin images. Next, we extend the computational method to generate archival-quality images in minutes, which are equivalent to those obtained from hours/days-long formalin-fixed, paraffin-embedded processing. We assessed the quality of images from the perspective of enabling pathologists to make decisions, demonstrating that the virtual stained image quality was diagnostically useful and the interpathologist agreement on prostate cancer grade was not impacted. Finally, because this method does not wash away lipids and small molecules, we assessed the utility of lipid chemical composition in determining grade. Together, the combination of chemical imaging and AI provides novel capabilities for rapid assessments in pathology by reducing the complexity and burden of current workflows. Significance: Archival-quality (formalin-fixed paraffin-embedded), thin-section diagnostic images are obtained from thick-cut, fresh-frozen prostate tissues without dyes or stains to expedite cancer histopathology by combining SRS microscopy and machine learning.