Abstract 2065: A deep-learning approach for tissue mapping and cell phenotyping in imaging mass cytometry data analysis

Abstract Imaging Mass Cytometry™ (IMC™) technology is a multiplexed imaging technique that generates high-dimensional spatial data at subcellular resolution without the complications of autofluorescence and cyclic imaging. IMC technology has two distinct whole slide imaging (WSI) modes: Preview Mode (PM) and Tissue Mode (TM). PM rapidly scans stained tissue to provide a comprehensive overview within minutes, while TM provides fast acquisition of the entire tissue at 5-micron resolution, mapping out the distribution of over 40 markers and revealing tissue heterogeneity. Both WSI modes enable researchers to make informed decisions about selecting tissue areas that warrant closer examination at single-cell resolution. Following PM, regions of interest (ROIs) are selected on the same slide for high-resolution imaging using Cell Mode (CM). This facilitates single-cell analysis of the ROIs identified during PM. These imaging modes together with an automated slide loader function support non-stop acquisition of tissue samples all week. Tissue sections of lung cancer were stained with a 40-marker IMC panel by combining the Human Immuno-Oncology IMC Panel with the Human Immune Cell Expansion IMC Panel to study spatial organization and cellular interactions in the tissue. Images were acquired on the Hyperion™ XTi Imaging System (Standard BioTools), first in PM and then in CM with automatic selection of ROIs using Phenoplex™ software (Visiopharm®). ROIs were automatically selected based on three criteria: 1) tertiary lymphoid structures (TLS) expressing CD20 and CD3; 2) Granzyme B-rich areas and 3) areas with high number of CD68 and Vimentin double-positive cell clusters. An adjacent serial section was acquired in TM for a whole slide morphologic segmentation comparison. Tissue segmentation for all modes was performed using Phenoplex by training a deep-learning algorithm to recognize morphological features of the images, including the TLS. Single-cell analysis of the images generated in CM was performed using Phenoplex; cell segmentation was based on iridium DNA channels, and phenotyping was performed using the guided workflow. This data was used to compare the immune contexture through a series of t-SNE plots partitioned by spatial region and clinical variables. Many mature TLS with proliferating germinal centers were present in the tissue. A high number of granzyme B-expressing neutrophils were found. Vimentin-positive macrophage-like cells were associated with the TLS. This work demonstrates that the Hyperion XTi Imaging System with its three acquisition modes (PM, TM and CM) can greatly advance the ability of IMC users to obtain answers from complex samples. The interactive capabilities of Phenoplex allow the user to quickly identify pertinent cell types, find them within a tissue map, define their spatial relationship and analyze their neighbors, leading to valuable biological insights. Citation Format: Smriti Kala, Katherine Hales, David Mason, Hiroyuki Suzuki, James Robert Mansfield, Regan Baird. A deep-learning approach for tissue mapping and cell phenotyping in imaging mass cytometry data analysis [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 2065.