Intraoperative margin assessment (IMA) is critical in surgical pathology, especially related to vital organs, such as the human brain, which is a crucial part of the central nervous system. Frozen section (FS) is the global standard for IMA, which involves cryosectioning, susceptible to artifacts, consumes up-to 30 minutes per round, and eventually limits the number of IMAs in a critical surgery, which often leads to recurrence of a malignant tumor owing to leftover positive margins. Therefore, a reliable high-speed IMA is always a clear necessity in different levels of surgical pathology applications.

In the paradigm of modern digital IMA, for a fresh unsectioned tissue, it is often expected to collect ample amounts of information during the IMA itself. The purpose is to digitally secure decent diagnostic reliability so that the specimen does not require to be fixed or deep-frozen and physically preserved for a subsequent re-assessment. Several optical imaging modalities attempted to provide a rapid assessment with optical virtual sectioning of a specimen. Nevertheless, such prior digital IMA-potential studies often might not meet the prevailing state-of-the-art whole-slide-imaging (WSI) standard in the context of either a Nyquist-satisfied centimeter-scale gigapixel-sampled half-a-micron digital resolution, and/or a real-time artifact-free stitching/mosaicking feature, and/or a subminute gigapixel acquisition and digital display ability with an uncompromised resolution across the “specimen to digital display” pixel pathway.

Aside from speed and resolution concerns, the prior FS-alternative IMA approaches often encountered an accuracy/reliability concern. It is noted that pathologists across the globe are trained and adapted to the standard hematoxylin & eosin (H&E)-specific histological features. While adapting to H&E-alternative dyes, especially a source of nuclei contrast other than the gold standard, it is obvious that the results/images produced would not be readily assessed by a pathologist. That eventually needs dedicated deep learning and/or additional interpretation training for a pathologist, while establishing the same degree of accuracy and reliability as the gold standard remains another concern, not to mention repeated organ-specific and/or even hospital-specific training might be required for different surgical pathology applications.

In this research, we introduce an FS-alternative digital IMA-capable technique called Training-Free True-H&E Rapid Fresh digital-Pathology (the-RFP) that provides 4× faster assessment compared to FS-biopsy, yet holds an excellent reliability comparable to a formalin-fixed paraffin-embedded (FFPE) biopsy. With H&E compatibility, the-RFP requires no additional interpretation training for a pathologist. Unlike an FS-biopsy, the-RFP is free from tissue-freezing and physical sectioning, and thus does not encounter typical frozen-section artifacts. The-RFP is assisted by a mesoscale Nonlinear Optical Gigascope (mNLOG), which provides a truly WSI-competing whole specimen superficial imaging (WSSI) digital IMA-potential solution with an ability of multichannel imaging of a 1 cm² area in <120 seconds with 86 G bits or 3.6 Gigapixels of data, yielding a sustained effective throughput of >700 M bits/s, preserving a submicron digital resolution with no post-acquisition image processing. mNLOG platform is streamlined to a rapid artifact-compensated 2D large-field mosaic-stitching (rac2D-LMS) approach yielding a mosaic-stitching ability of >12×12 mm² area with 130 G bits of data in 60 seconds.

In this study, the-RFP is applied to investigate human brain tumors, being one of the fatal cancers across the globe. Despite unprecedented revolutions in medical sciences and technologies, the 5-year survival for glioblastoma cases only improved from 4% to 7% within the period 1975-1977 to 2009-2015, revealing its extreme seriousness. A fast and reliable IMA is thus critical to effectively resect such a fatal tumor while not much injuring eloquent regions to help circumvent postoperative neurological complications. A non-interventional clinical trial is conducted in collaboration with the Division of Neurosurgery, Department of Surgery, and Department of Pathology of National Taiwan University Hospital to investigate the accuracy and reliability of the-RFP technique. With no additional interpretation training provided to our collaborating pathologists, blind tests reveal an excellent accuracy of 100% considering a total of 50 normal and tumor-specific human brain specimens, where the-RFP-based decisions matched the respective FFPE-pathology outcomes.