A computer programme will learn to recognise normal tissue in pathology samples, freeing up specialist doctors to focus on complex cancer cases. Currently, pathologists spend a large proportion of their time examining microscope slides that turn out to be normal. This is expensive and inefficient, and it diverts expert attention from difficult diagnoses where human judgment is most needed. At the same time, pathologists’ performance can vary, especially for complex tasks in breast and prostate cancer, meaning some patients do not receive the best treatment. If successful, PathLEAD will produce two practical tools: one that automatically flags normal samples so they need no human review, and another that assists pathologists with challenging cases to improve diagnostic accuracy. The consortium—spanning Coventry, Belfast, Nottingham, Oxford, and Philips—will also create a large repository of scanned pathology images, made available to other UK researchers and companies. This shared resource could accelerate development of further diagnostic tools across the NHS. Patients can opt out of data use via the NHS National Data Opt Out scheme, and a lay committee will oversee ethical decisions. Any commercial income will be reinvested into the NHS.
View original technical description
"PathLEAD (**Path**ology data **L**akes, **E**ducation, **A**nalytics and **D**iscovery) is a consortium of nationally leading experts from the teaching hospitals of Coventry, Belfast, Nottingham and Oxford, their associated Universities and Philips, the commercial partner. We will develop computer aided diagnostics for testing of pathology samples. Specialist doctors (called pathologists) currently carry out testing via visual examination of pathology specimens under the microscope, a process that is inherently subjective. A large proportion of tissue samples examined are normal and using specialist pathologist time to establish this is expensive. We believe this can be done by computers, and valuable pathologist time saved and used elsewhere. Therefore we will develop a computer programme that will recognise normal tissue so that the sample does not need to examined by the pathologist. Furthermore we know that pathologists' performance is variable and in some cases limited, particularly where the tasks they perform are complex or require extensive experience. This can mean patients with some forms of breast and prostate cancer do not get the best treatment. Our computer programmes will assist the pathologist and improve these decisions. The development of these tools requires thousands of image files obtained from scanning microscope slides. This is time consuming to collect so once this data has been produced it is a valuable resource. We aim to make this available to other research teams, our commercial partner Philips and UK-based companies developing tools to improve healthcare. The result will be greater knowledge, improved tools and better care for the future. We recognise that patients have the right not to allow their data to be used for this purpose, and will be using the recently launched NHS National Data Opt Out scheme to record patients wishes accurately. Patients and lay representatives on our ethical and management committee will help decide how this data should be used. We expect that some tools will become successful commercially and plan to exploit this success re-investing some of the income back into the NHS to benefit patients. Our ability to provide high quality data, expertise, access to top-grade computer equipment and knowledge of commercialising these tools, will percolate to UK companies, building the economy in this sector and lead to the UK becoming a global leader. Our expertise will provide education to the pathology and computer science communities to share the knowledge gained, gaining in quality and efficiency, and having patients' benefit as our ultimate goal."
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