Lung cancer: How AI can be used to improve treatment management
Oliver Humbert, doctor and professor of nuclear medicine and biophysics at Cote d’Azur University, explains AI’s uses in treating lung cancer.
Cancer is one of the primary causes of death in the UK: one in two people will be diagnosed with the disease in their lifetime. Lung cancer is the third most common cancer type in the UK, with 47,838 new cases diagnosed between 2015-2017.
Recent therapeutic innovations have increased life expectancy and quality for cancer patients, even at an advanced stage of the disease. However, there is still progress that can be made in providing more personalised oncology; individualising therapeutic strategy according to the biological characteristics of the disease for each patient - to improve effectiveness and limit the side effects of treatments. Technology, and particularly artificial intelligence, can be a valuable asset in improving the care of patients with advanced lung cancer.
Immunotherapy, a treatment which has proven itself but also has limitations
Immunotherapy is a therapeutic tool that works by reactivating the patient’s immune system, rather than targeting cancerous cells directly. It stimulates the patient’s lymphocytes (immune cells) to reinstate their ability to recognise and destroy cancerous cells. This treatment has achieved very good results - even when treating metastatic lung cancer. However, not all patients respond well. Sometimes the cancer finds a means of escape, preventing the patient’s receptiveness to immunotherapy. The treatment can also cause inflammatory side-effects.
So how do you know whether immunotherapy will be a more effective treatment than another chemotherapeutic choice in fighting the disease? Over the next few years, one of the main goals in the fight against cancer is to understand how to predict which patient will respond to immunotherapy. This will enable doctors to target patients that should be offered the treatment.
That goal will be achieved thanks to artificial intelligence.
Predict to cure: technology to serve doctors
Artificial intelligence is based on deep and machine learning. More specifically, the technology relies on the collection and analysis of large volumes of data to make predictions. This is done almost instantaneously, something which is not humanly possible.
In terms of cancer, the progress of the disease is monitored via a PET scan. A benchmark in oncology, it allows the activity of tumour cells to be measured through an intravenous injection of a mildly radioactive glucose derivative. The PET scan also takes precise images of the patient’s anatomy, the morphology of their organs and the cancer itself. This examination, therefore, makes it possible to gather large volumes of digital data about the patient and their disease, most of which is hidden in the depth of the image and cannot be analysed by the human eye.
At Sophia Antipolis, the Institut 3IA has set up a research project to apply artificial intelligence to the analysis of PET scan data from lung cancer patients treated with immunotherapy. Using an algorithm, thousands of data items contained within the patient’s image are accurately analysed, in 3D, from both a morphological and biological perspective, at the commencement of treatment and thereafter at two-monthly intervals.
After an initial phase of learning, the algorithm can begin to identify features of the image that relate to the response to immunotherapy. This builds a model that is then capable of predicting whether the treatment will be effective or not, for any given patient. The more the algorithm is trained on many patients and PET scan images, the more it will learn and the better its final performance will be.
Longer term, the theory is that the image of a patient can be analysed using the algorithm before any treatment is decided. This will provide a probable success or failure rate for immunotherapy, for each patient – enabling doctors to tailor the therapeutic strategy for them. In this way, artificial intelligence becomes a tool to help doctors in their decision-making process and represents a real opportunity to improve cancer treatment.
But, as with any tool, it is necessary for the medical profession to master it, and to understand its powers and limitations. That’s why graduate training programmes are being developed, like the French post graduate course – the Diplome Universitaire IA et Santé, to support doctors in this process.
Be part of the medicine of tomorrow
This new way of practicing medicine allows silos that may exist between mathematicians, doctors, and commercial health organisations to be decompartmentalised. It also allows them to collaborate and define algorithms and other innovative AI solutions for patients.
It is only through multidisciplinary collaborations like this that researchers and health professionals can come together to develop the medicine of tomorrow.