Every year, thousands of NHS colorectal cancer patients wait an average of 14 days for a test result that could be delivered in 24 to 48 hours. The problem is that two essential tissue-based tests—MSI/MMR for colorectal cancer and PD-L1 for lung cancer—are slow, expensive, and unevenly available across the UK. One relies on a pathologist’s subjective interpretation, creating inconsistency. The other uses costly PCR or immunohistochemistry methods that could be replaced by analysing a standard H&E tissue section. This project aims to replace those approaches with artificial intelligence. The AI would objectively score PD-L1 staining and predict MSI status directly from routine microscope slides, eliminating the need for separate molecular tests. A centralised digital platform would let NHS labs upload scanned images and receive results within two days, rather than two weeks. If successful, the system could make tissue-based molecular testing faster, cheaper, and more consistent across the country. It would also create a delivery infrastructure ready for future AI diagnostics in the NHS. The research is applied and practical—designed for immediate clinical deployment, not fundamental discovery.
View original technical description
Background The 43,000 NHS colorectal cancer (CRC) patients diagnosed yearly require microsatellite instability or mismatch repair testing (MSI/MMR), at a cost of between £4,085,000 and £8,385,000. Approximately 36,000 patients with non-small cell carcinoma (NSCLC) require PD-L1 testing, costing £7M. MSI results take an average of 14 days and PD-L1 delivery is difficult, being based on a pathologist’s subjective interpretation. There are geographical gaps across the UK in the delivery of these necessary tests. Aims Our proposal is based on two key pillars: 1. Bringing Artificial Intelligence to routine diagnostics, applied to: a) the basic H&E section, in MSI/MMR, replacing the current PCR/IHC approaches b) objective scoring of PD-L1 IHC. 2. Creating a centralized platform: a) for submission of in-silico images and b) delivery of such algorithms across the NHS. Work Plan We propose work packages to: a) develop and validate the AI algorithms (ImageDx Lung/Colon) on retrospective and prospective datasets. b) create a delivery platform (Sonrai/PMC Central Hub, SPCH) under ISO13485. c) achieve clinical validation and regulatory approval. This novel modality applying AI to diagnostics requires clear explanation to patients, and patient input will be decisive to delivery; a detailed PPI plan is put forward to capture this important contribution. Dissemination and Anticipated Outcomes Dissemination will happen in three phases: a) Northern Ireland. b) PathLAKE membership. c) NHS wide. Our approach will create direct annual savings across the NHS to the tune of £648,500 – £3,228,500. Indirect savings associated with anti-PDL1 therapy based on more accurate testing could impact 10% of the patients and on £35M of the budget on therapeutics. Following the NHS drive towards testing centralization, the SPCH will receive online scanned images and render results in 24-48 hours. This model of AI, tissue-based molecular diagnostic delivery will enable adoption of future AI- based testing in the NHS.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know