TLS_Finder: An algorithm for Identifying Tertiary Lymphoid Structures Using Immune Cell Spatial Coordinates
Year: 2024
Type: preprint
Abstract: Tertiary lymphoid structures (TLS) are lymphoid formations that develop in non-lymphoid tissues during chronic inflammation, autoimmune diseases, and cancer. Accurate identification and quantification of TLS in tissue can provide crucial insights into the immune response of several disease processes including antitumor immune response. TLS are defined as aggregates of T cells, B cells and dendritic cells. In histological tissue sections stained with Hematoxylin and Eosin they are identified as aggregates of 50 or more lymphoid cells, however immunohistochemical analysis are required to confirm presence of distinct immune cell patterns. Assessment of lymphoid aggregates can be done in H&E slides or in slides stained with single or multiplex immunohistochemistry or other tissue based high-plex approaches with key biomarkers such as CD3 (T cells) and CD20 (B cells), however manual assessment of them is time consuming thus limiting its full evaluation. To our knowledge, published algorithms that identify TLS within a tissue are based on histological assessment of H&E slides or through use of ML algorithms trained on images that show the TLS presence in tissues; and quantification and spatial analysis of TLS still remains a challenge. This study aims to develop a robust algorithm to recognize TLS using the spatial coordinates of immune cells in any given tissue. The algorithm uses X and Y coordinates of T and B cells in tissues identified by a pathologist-supervised digital image analysis of multiplex chromogenic immunohistochemistry of tissues stained with CD3 and CD20. The algorithm is flexible to be used for detailed analysis of TLS stages; including other cell types within the definition of TLS; such as dendritic cells (DC) and high endothelial venules (HEV); to assess different stages of TLS formation.
Authors: Ayse A Koksoy, Maria Esther Salvatierra, MOSAIC Team, Luisa Maren Solis Soto, Cara Haymaker
Institution: The University of Texas MD Anderson Cancer Center
Language: English
Topic: AI in cancer detection
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4405851240