A revolutionary artificial intelligence model developed by the Octopath AI team is set to transform cancer diagnostics and biomarker discovery, offering unmatched accuracy and significantly reducing patient wait times.
Cancer diagnosis is currently bottlenecked by a 25% shortfall in UK pathology staff and a projected 77% global rise in cancer cases by 2050. This strain causes critical delays in diagnostic pathways, impacting patient care and incurring unnecessary costs to the NHS.
To address this critical bottleneck, the founding team behind University College London (UCL) spinout Octopath AI – Dr Zhuoyan Shen, Dr Charles-Antoine Collins-Fekete, and Dr Esther Baer - developed a revolutionary solution. Their new foundation cell model, Kraken-1, transforms computational pathology by reading tissue at single-cell resolution. While traditional AI relies on slow, biased, hand-drawn manual annotations, Kraken-1 is built on an automatically generated database of 14.1 million labelled cells using a proprietary method.
The Kraken-1 solution was developed to provide a series of highly accurate bespoke models that automate mandated pathologists' tasks, freeing their time so they can commit themselves where they bring the highest value.
The dataset spans 14 different tissue and cancer types, including breast, lung, colon, brain, and liver, and allows the model to generalise without requiring per-organ retraining. The result is a 30% higher accuracy compared to the previous state-of-the-art solutions.
Kraken-1 outperforms leading commercial and academic single-task AI models. Benchmarks show the model achieving 86% accuracy in pan-cancer mitosis detection and 71.1% accuracy in cell-type classification. In a single pass, the model can localise and classify over ten cell types, including epithelial, lymphocyte, myeloid, fibroblast, vascular, and smooth muscle cells.
The Octopath team claims that acting as an AI companion, Kraken-1 turns routine biopsies into precision medicine insights in under two minutes. The unique foundational architecture means that new diagnostic models can be launched in a matter of weeks, a significant improvement over the months or years currently offered.
The technology translates pixels into reportable clinical metrics out of the box, powering tumour cellularity scoring, Ki67 proliferation indices, tumour-stroma ratios, and tumour budding assessments.
The deployment of Kraken-1 offers transformative benefits for stakeholders across the medical field:
- For patients: the platform can shorten diagnostic waiting times by an average of four days, enabling faster access to personalised disease control
- For pathologists: the system eliminates the burden of manual cell counting, reducing case review time by 20% to 55%, minimising fatigue, and freeing up capacity for complex cases
- For healthcare systems: increased efficiency and reduced outsourcing fees offer potential health economic savings of £100,000 to £200,000 annually per trust
- For the pharmaceutical industry: the model offers a deployable tool to accurately quantify therapy-related biomarkers across global clinical trials, directly accelerating new cancer therapies.
Identifying new cancer biomarkers often requires expensive and technically demanding molecular profiling techniques, such as single-cell sequencing or spatial transcriptomics, which are costly to deploy at a population scale. Kraken-1 enables researchers to overcome this barrier by extracting biologically and clinically meaningful information directly from routinely collected pathology slides. This provides a lower-cost and more scalable approach to biomarker discovery, creating opportunities to identify new prognostic and predictive biomarkers across large patient cohorts and accelerating both cancer research and precision medicine.
Kraken-1 is optimised for clinical deployment without cloud and all data remaining securely contained on the premises. A free version is accessible upon request on the Octopath web platform, www.octopath.ai.