Octopath AI, a UCL spinout developing artificial intelligence for digital pathology, has launched with £1.5 million in combined seed investment and grant funding, led by TwinPath Ventures.
The company is building tools designed to help pathologists draw more information from each routine tissue slide, at a time when cancer services are under sustained pressure.
Built on research from University College London’s (UCL) Department of Medical Physics and Biomedical Engineering, Octopath AI is developing AI designed to work alongside pathologists, analysing routine tissue slides, both haematoxylin and eosin (H&E) and immunohistochemistry (IHC), at single-cell resolution in under two minutes. A pathologist reviewing the same slide by eye would not attempt to characterise every cell individually. The platform instead returns quantitative measurements, such as cell counts, densities and spatial relationships, that would otherwise never be captured from the slide at all. By comparison, manual estimation of tumour cell density takes a pathologist around 20 minutes per slide.
The launch comes at the start of National Pathology Week (2-9 November) and as pathology services face rising demand and workforce pressure. In May 2026, more than 55,000 people in England waited longer than the NHS’s four-week standard to find out whether they had cancer, while only 3% of UK histopathology departments report having enough staff to meet clinical demand. Delays matter: a four-week delay to adjuvant chemotherapy for colorectal cancer has been associated with a 13% increase in mortality risk.
Octopath AI’s platform is built on two foundation models. Kraken analyses routine H&E slides, counting and classifying cells and tissue structures, and has been trained on 14.1 million individually labelled cells across 14 cancer types. Nemo analyses immunohistochemistry (IHC), scoring every cell on a slide for stain positivity and intensity. Because an IHC score is fundamentally a ratio of cells, counting all of them rather than a sampled few hundred removes much of the sampling variation from the result.
The models are trained using ‘spatial-omics’ data – measurement of biological molecules such as RNA, proteins or metabolites, while preserving exactly where those molecules sit within a tissue.
The platform is designed to support a range of analytical applications spanning mitosis detection and tumour grading, immune-cell mapping, cellularity, blood-vessel analysis and biomarker scoring, including PD-L1, Ki-67, ER/PR and HER2. By producing multiple readouts from a single platform, Octopath AI aims to support pathology services, pharmaceutical companies and clinical-trial teams in generating the quantitative evidence that underpins research into targeted treatments.
The approach has been evaluated in two peer-reviewed studies published in eBioMedicine. In an analysis of 414 biopsies from the phase 3 ARISTOTLE rectal cancer trial, an AI-derived measure of tumour cell density identified a subgroup of patients whose outcomes differed significantly depending on which chemoradiotherapy regimen they received, a difference not visible in the trial overall.
A companion study found that AI-measured immune cell densities were associated with survival, and improved prognostic accuracy when added to standard clinical variables, with validation in independent cohorts from University College London Hospitals (UCLH) and The Cancer Genome Atlas.
Dr Zhuoyan Shen, co-founder and Chief Scientific Officer of Octopath AI and first author of both studies, said: “Measuring a slide by sampling a few regions inherits all the variability of that choice; counting every cell removes it. In our ARISTOTLE analysis, that precision surfaced differences between patient groups that the trial as a whole could not see. The point is not that the AI is faster; it is that it measures things that were never measurable by eye.”
Octopath AI’s technology is in use at five sites, including the Division of Neuropathology at the National Hospital for Neurology and Neurosurgery, part of UCLH, where it processes around 200 slides a day. There, the software counts dividing cells and measures other features in brain tumour samples, helping neuropathology specialist doctors save time and judge how quickly a tumour may grow.
Samples from four clinical trials are currently in analysis, including SCOPE2 and PET-NECK. The company’s methods have also been applied to archived biopsies from the completed phase 3 ARISTOTLE trial.
Octopath AI is one of seven companies selected for the MHRA’s AI Airlock Phase 2, the regulator’s sandbox for AI medical devices, and has secured a £500,000 Innovate UK Growth Catalyst grant. The grant is unlocked by the seed round on a matched basis, bringing total funding for this stage to £1.5 million. The funding will support clinical evaluation, regulatory and quality-management work, product development and commercial activity as the company works towards regulatory approval.
Octopath AI was founded by Dr Zhuoyan Shen, Dr Charles-Antoine Collins Fekete and Dr Esther Bär, all of UCL’s Department of Medical Physics and Biomedical Engineering. The company was established as a spinout with support from UCL Ventures, the commercialisation company for UCL, which worked with the founders to protect the underlying research, identify the right market fit and connect them to investors.