Our tissue segmentation algorithm uses a combination of U-Net architecture and attention mechanisms to accurately delineate various tissue types within complex histological images. This segmentation allows for more precise quantification of tissue composition and can reveal subtle changes in tissue structure.\n\nThe algorithm has been trained on a diverse set of tissue samples and has demonstrated high accuracy in distinguishing between different tissue types, even in challenging cases with complex morphologies.
20 February 2025 · 1 min read
Tumour Stroma Ratio
Learn how our tissue segmentation algorithm accurately delineates different tissue types, from stroma to tumor regions
Charles-Antoine Collins-FeketeOctopath
Share this post
Charles-Antoine Collins-Fekete
Dr. Collins-Fekete is a UKRI Future Leaders Fellow at UCL, leading research in AI for cancer diagnosis with a focus on digital pathology. He has established a team of post-doctoral researchers, published over 30 peer-reviewed papers, and secured substantial funding exceeding £3 Mio. As founder of the Octopath spin-out and co-founder of the UCL Cancer Collaboratorium, he drives the translation of cutting-edge science into impactful solutions for cancer care.
More
Also from the blog
Algorithms28 October 2025 · 1 min read
Atypical Mitoses
An atypical mitosis is an abnormal cell division, a key hallmark of cancer. Its features, like multipolar spindles, are crucial for grading a tumor's aggressiveness and predicting prognosis.
Charles-Antoine Collins-Fekete
Algorithms27 October 2025 · 1 min read
Tumour Cell Density
Our latest methodogy to separate tumour cells from background cells and calculate density in colorectal cancer.
Zhuoyan Shen
Algorithms23 October 2025 · 1 min read
Cytoprofile
AI model analyzes standard H&E slides to identify cell types and map tumor microenvironments, bringing spatial-omics insights to routine pathology without the cost.
Esther Baer