New study to help predict metastatic melanoma patient survival

Researchers in Medicine are to begin a new project with University of Southampton computer scientists using clinical data with machine learning to predict cancer patients’ survival.

Uveal melanoma is the most common primary eye tumour in adults but has a poor prognosis with only one in 10 patients surviving after one year. University Hospital Southampton specialises in providing a promising minimally invasive treatment called Melphalan Percutaneous Hepatic Perfusion. This involves delivering chemotherapy directly to the tumour arteries via a small hole in the groin. This has shown great promise, but outcomes vary significantly.

The new study, funded by an Academy of Medical Sciences Starter Grant for Clinical Lecturers, will use machine learning to help predict which patients will respond to treatment.

Dr Ganesh Vigneswaran, NIHR Clinical Lecturer in Cancer sciences, will lead the study. He explained: “While treatments for cancer are helping many people, we cannot predict who will respond well to treatment.

“Machine learning is a type of artificial intelligence that can find complicated patterns in big datasets, such as treatment response or survival outcomes. Most patients have routine CT scans (reconstructed detailed images of the body) and these scans might contain important clues specific to each patient. Our goal is to employ artificial intelligence to extract and uncover this information to help predict treatment responses. If we can establish which patients are likely to respond and which treatments are likely to be useful, we can improve decision making and save patients from ineffective treatment and side effects.”

The study will also develop a tool to aid in shared decision-making for metastatic uveal melanoma patients to improve current treatment paths.

Furthermore, the approaches used should be transferrable to other ‘image-centric’ therapies, such as TACE (trans-arterial-chemoembolization) and SIRT (Selective-internal-radiation-therapy), both of which also involve cancer treatment via blood vessels. This could serve as the foundation for the use of routine imaging in predicting patient outcomes in many other cancer types.