Multi-resolution open-top light-sheet microscopy to enable 3D pathology of lymph nodes for breast cancer staging

Abstract

Slide-based histopathology is the clinical gold-standard technique used to diagnose cancer and determine a patient’s prognosis. Unfortunately, its basis in sparse 2D visualization and tedious, destructive tissue processing reduces diagnostic accuracy, reproducibility, and inter-observer agreement, potentially compromising the quality of patient treatment. To address these limitations, open-top light-sheet microscopy (OTLS) has been developed for rapid, nondestructive 3D imaging of large pathology specimens. It offers the same level of detail as traditional pathology slides and has been applied to various pathology applications such as breast and prostate cancer. In this work, we overcome several limitations of previous OTLS architectures such as limited imaging depth and an inability to accommodate various clearing protocols with development of a solid immersion meniscus lens (SIMlens). A SIMlens is a wavefront-matching element that enables the use of air-based objectives, which, when implemented with a turret in OTLS microscopy, allows users to rapidly transition between low- and high-resolution 3D views. We report development of the first multi-resolution OTLS microscope based on this SIMlens technology, which enables efficient 3D imaging in diagnostic pathology applications. Additionally, we explore how multi-resolution OTLS microscopy may facilitate improved management of breast cancer patients. We showcase a comprehensive 3D pathology workflow for staging metastases in whole lymph nodes with multi-resolution OTLS microscopy, and demonstrate that this workflow may improve breast cancer staging accuracy in comparison to standard of care. Finally, to further streamline diagnosis with 3D pathology, we implement a deep learning-based computational triage method to automatically segment suspicious regions in 3D pathology datasets to guide pathologist review. We demonstrate this method with screening and diagnosis of dysplasia and esophageal adenocarcinoma (EAC) in endoscopic biopsies from patients with Barrett’s esophagus (a precursor to EAC). In future work we aim to show that our AI-assisted 3D pathology method may improve diagnostic sensitivity of neoplasia in endoscopic biopsies while reducing the amount of information that pathologists must review. Importantly, this could facilitate earlier detection and treatment of EAC.