AI tool accurately predicts tumour regrowth in cancer patients
Doctors and scientists have developed an artificial intelligence tool that can accurately predict how likely tumours are to grow back in cancer patients after they have undergone treatment.
The breakthrough, described as “exciting” by clinical oncologists, could revolutionise the surveillance of patients. While treatment advances in recent years have boosted survival chances, there remains a risk that the disease might come back.
Monitoring patients after treatment is vital to ensuring any cancer recurrence is acted on urgently. Currently, however, doctors tend to have to rely on traditional methods, including ones focused on the original amount and spread of cancer, to predict how a patient might fare in future.
Now a world-first study by the Royal Marsden NHS Foundation Trust, the Institute of Cancer Research, London, and Imperial College London has identified a model using machine-learning – a type of AI – that can predict the risk of cancer coming back, and do it better than existing methods.
“This is an important step forward in being able to use AI to understand which patients are at highest risk of cancer recurrence, and to detect this relapse sooner so that re-treatment can be more effective,” said Dr Richard Lee, a consultant physician in respiratory medicine and early diagnosis at the Royal Marsden NHS Foundation Trust.
Lee, the chief investigator of the OCTAPUS-AI study, told the Guardian it could prove vital in not only improving outcomes for cancer patients, but alleviating their fears, with relapse “a key source of anxiety” for many. “We hope to push boundaries to improve the care of cancer patients, to help them live longer, and reduce the impact the disease has on their lives.”
The AI tool may lead to recurrence being detected earlier in patients deemed at high risk, ensuring they receive treatment more urgently, but it could also result in fewer unnecessary follow-up scans and hospital visits for those deemed low risk.
“Reducing the number of scans needed in this setting can be helpful, and also reduce radiation exposure, hospital visits, and make more efficient use of valuable NHS resources,” Lee said.
In the retrospective study, doctors, scientists and researchers developed a machine learning model to determine whether it could accurately identify non-small cell lung cancer (NSCLC) patients at risk of recurrence following radiotherapy. Machine learning is a form of AI that enables software to automatically predict outcomes.
Lung cancer is the leading worldwide cause of cancer death and accounts for just over a fifth (21%) of cancer deaths in the UK. NSCLC makes up nearly five sixths (85%) of lung cancer cases and, when caught early, the disease is often curable. However, over a third (36%) of NSCLC patients experience recurrence in the UK.
The researchers used clinical data from 657 NSCLC patients treated at five UK hospitals to feed their model – and added in data on various prognostic factors to better predict a patient’s chance of recurrence.
These included the patient’s age, gender, BMI, smoking status, the intensity of radiotherapy, and their tumour’s characteristics. Researchers then used the AI model to categorise patients into low and high risk of recurrence, how long a period they might experience before a recurrence, and overall survival two years post treatment.
The tool was found to be more accurate in predicting outcomes than traditional methods. The results of the study, supported by the Royal Marsden Cancer Charity and the National Institute for Health Research, were published in The Lancet’s eBioMedicine journal.
“Right now, there is no set framework for the surveillance of non-small cell lung cancer patients following radiotherapy treatment in the UK,” said study lead Dr Sumeet Hindocha, a clinical oncology specialist registrar at the Royal Marsden and Imperial College London. “This means there is variation in the type and frequency of follow-up that patients receive … Using AI with healthcare data may be the answer.
“As this type of data can be accessed easily, this methodology could be replicated across different health systems.”
The study is “an exciting first step” towards rolling out a tool nationally and internationally to guide the post-treatment surveillance of cancer patients, Hindocha added.