CRUK, together with the Science and Technology Facilities Council, have organised a series of sandpits to foster multidisciplinary research collaborations focussing on developing innovative ideas for early detection. The latest workshop, held on 18-20th November 2019, concentrated on the application of artificial intelligence to the analysis of digital pathology images.
The majority of cancer diagnoses will require histological or cytological analysis, however, a shortage of pathologists means that these diagnoses can be delayed. Increasing analysis automation would decrease turnaround times, and may also improve diagnostic accuracy and the identification of earlier cancers or pre-cancers that may otherwise have been missed. This ambition comes with many challenges, including how to define the normal, pre-cancerous and cancerous states so that they are accurately distinguished by a computer algorithm, and how to deal with debris on slides and variability in sampling.
During the workshop, these challenges were discussed and the participants split into teams to devise projects aimed at tackling one or more of these problems. Four scientists from Oxford attended: Daniel Royston (Nuffield Division of Clinical Laboratory Sciences, RDM); Alistair Easton (Department of Oncology); Heba Sailem (Institute of Biomedical Engineering); and Korsuk Sirinukunwattana (Institute of Biomedical Engineering).
A project led by Daniel Royston, with team members Alistair Easton, Korsuk Sirinukunwattana, Wei Pang (University of Aberdeen), Fayyaz Minhas (University of Warwick), Peter Dunstan (Swansea University) and Matthew Grech-Sollars (Imperial College London) was successful in attracting £100,000 funding for pilot studies. Team Haem-AI will focus on the earlier diagnosis of haematological malignancies. One difficultly with diagnosing these malignancies is that they are often closely mimicked by inflammatory conditions and thus require sequential samples to accurately differentiate between these states. The team propose to utilise artificial intelligence to distinguish between these states at an earlier time-point in the context of myeloproliferative neoplasms (MPN) and mycosis fungoides.
Heba Sailem is a member of the two other successful teams. Team PathNAV, led by Marnix Jansen (University College London), is aiming to detect pre-neoplastic changes in Barrett’s oesophagus earlier and team PRISM, led by Jan Lukas Robertus (Imperial College London), is focussing on machine learning approaches to mesothelioma pre-neoplastic signature discovery.
The next early detection innovation sandpit is themed on novel technological approaches for the earlier detection of pancreatic cancer. For more information and how to apply, visit the CRUK website.