AI-powered innovation gives new hope for treating advanced lung cancer
CU Medicine develops an AI-powered analytic tool for lung cancer immunotherapy.
A pioneering analytic tool for predicting patient outcomes of an increasingly popular treatment for advanced lung cancer has been developed by an international team including researchers from CU Medicine and South Korean institutes. Using AI and deep learning, the technique can help doctors choose the best clinical interventions for patients and potentially improve survival rates. The findings were recently published in the Journal of Clinical Oncology.
Lung cancer is one of the most common cancers in the world, accounting for 1.8 million deaths every year. A variant known as non-small-cell lung cancer accounts for more than 80% of all lung cancers. Recently, oncologists have achieved promising results by treating non-small-cell lung cancer through immunotherapy, an approach based on supercharging the immune system’s own natural capacity to fight cancer, as an alternative to traditional chemotherapy.
Predicting performance of immune interventions
In some cases, for example, immunotherapeutic drugs can adjust the immune system’s self-regulating mechanisms so that it can attack cancer cells more aggressively. The body contains immune cells that act as natural ‘checkpoints’ to our immune responses. Normally, these checkpoints stop our immune reactions from becoming too strong and killing healthy cells, but they can also prevent the body from destroying a cancerous tumour. Drugs that inhibit these immune checkpoints clear the blockage and allow the immune system to do its job.
Prof. Tony Mok of the Department of Clinical Oncology at CU Medicine contributed to the design of the new analytical tool which has significantly improved doctors’ ability to use a particular biomarker to predict how well this type of intervention will work given the condition of the patient’s tumour.
There is no standard biomarker for predicting how well such interventions will work. However, one promising indicator is the presence of certain white blood cells that infiltrate the tumour and activate the immune response. The problem is that counting these cells is currently very labour-intensive and relies on oncologists examining digital images of a whole microscope slide.
Prof. Tony Mok is a world-leading expert in lung cancer research.
Driving better clinical decisions
By contrast, the AI-powered tool rapidly analyses the spatial distribution of different parts of the tumour from whole-slide images, segmenting and quantifying its various components—including the tumour-infiltrating white blood cells that can act as a biomarker. Based on deep learning-based AI model training, it then categorises the immune status of the cancer. In some cases, there will either be no immune response at all, or an immune response that can’t penetrate the tumour micro-environment. However, there will sometimes be a potentially cancer-beating immune response within the tumour that is being held back by the body’s immune checkpoints. Such cases are the most promising candidates for interventions that inhibit those checkpoints.
By helping oncologists to extract more reliable information from this biomarker, the analytic tool gives a fuller picture of the patient’s condition and supports better clinical decision-making.
Prof. Mok collaborated with South Korea’s Seoul National University College of Medicine, Sungkyunkwan University School of Medicine, Ajou University School of Medicine and AI startup Lunit to develop the new tool.
Read more: