TissueReM: Whole-Organism, Real-time Decision-enabled 3D Tissue Imaging and Recovery for Molecular Analysis

Bioinvision, Inc., Cleveland OH

Investigators

Abstract

Summary. Traditional methods of studying tissue often limit researchers to individual dissected organs, hinder- ing a holistic, whole-organismal view. Whole-organism analysis provides a comprehensive picture of sys- temic effects of biology or disease on multiple organs and reveals vital insights about biological function. Applications are limitless including cellular therapy, gene expression studies, cancer biology and immunotherapy. To enable whole organism ex vivo analysis, we developed and commercialized CryoVizTM Imaging, which can serially section and image a whole mouse and can visualize high-resolution anatomy and multiple fluorescence signals, allowing for a three-dimensional (3D) visualization of disease from every angle with single-cell resolution. As the system sections through the organism, it also allows for recovery of the sections. However, it is very costly and time consuming to collect sections from the entire organism and at best, investigators might only guess where they want tissue sections based on their intuition about disease localization. But the real problem and hence the unmet need is that there is no existing method for a combined whole organis- mal level screening with targeted tissue collection. Noting the need for automatic tissue-focused investiga- tion and section collection to study the impact of disease, in Phase I, we created a baseline, AI-driven TissueReM software powered by CryoVizTM imaging, allowing for identification of a limited set of tissues. We achieved ≥80% reduction in operator supervision time and accurate tissue segmentation (DICE overlap>0.9), and sensitiv- ity/specificity ≥ 90%/90% with regards to human expert ground truth for 10 tissues. In Phase II, we will build upon TissueReM to create DL models to identify an expanded set of tissues (~20) (TID-Net-Seg), classify diseased versus heathy tissue (TID-Net-Path), and obtain “virtually stained” images from scanned im- ages of unstained tissue sections (VHS-Net). TID-Net-Seg will include glands in the head and neck region, lymphatic system (lymph nodes and lymph vessels), mesentery and associated structures, regions of the brain, pancreas, and adrenal glands. Noting the need for an organismal level screening for disease, we will create TID- Net-Path to differentiate between healthy and diseased tissue. Chemical staining is a major burden for histology labs, which requires the use and waste-processing of multiple types of costly and toxic reagents and antibodies. VHS-Net will create “virtually stained” H&E images from label-free scanned images of sections without needing chemicals and reagents. 3D block-face imaging data will be integrated with histology data through 3D rigid and non-rigid image registration, enabling correlation of histology at a cellular/sub-cellular level with block-face im- ages at the organ/tissue level. We will validate TissueReM in demonstration experiments spanning the applica- tion areas of invasion and metastases on nerves as well as CAR-T immunotherapy. For commercialization, we will identify target markets and offer TissueReM software platform as a fee-for-service product with all analysis, visualization, and images delivered to the customer.