A new research study has launched to progress how artificial intelligence technologies can help predict whether prostate cancer may return for patients following treatment.
Dr Anna Wilkins and Dr Erik Sahai, along with Prostate Cancer Research and partners, launched the study in June 2021, to explore whether computer modelling of cell migration can automatically highlight certain features around prostate cancer cells which may mean the cancer is more likely to return.
The researchers have developed the tool to help find these features and will now use the study to analyse tissue samples collected from patients with prostate cancer during its CHHiP clinical trial, and then further build models to classify prostate tumours based on different cell features.
The researchers said: “We’ve known for many years that specific patterns of non-cancerous cells, and the surrounding scaffolding that supports tumours, can drive more aggressive prostate cancer behaviour. However, we haven’t had the computational tools to find these “bad biology” tumour patterns.
“Recent advances in artificial intelligence mean there is now the potential to identify them. We’re very excited that our PCR award means we can apply powerful computational approaches to digital images of tumours, with the goal of enabling better treatment decisions for individual patients – maximising cure rates and minimising side effects.”
Following the study, the hope is for the tool to be implemented into clinics to support more personalised treatment for those diagnosed with prostate cancer. In the longer term, the aim is for the software to support different stages of prostate cancer to gain a better understanding of the biology that underpins aggressive disease and hormone therapy resistance.
Anna and Erik are working with other researchers from the Francis Crick Institute, the Institute of Cancer Research, Royal Marsden Hospitals NHS Trust and University College London. The team will also hire machine learning, statistics, and clinical trials experts into the study.
