Applying artificial intelligence to cancer diagnosis
A pioneering research program at the JGH is connecting the power of artificial intelligence (AI) and radiomics to heighten the effectiveness of cancer treatment by gleaning as much information as possible from radiological imaging technology.
The goal of the program, directed by Dr. Reza Forghani in the Segal Cancer Centre at the JGH, is to improve diagnostic accuracy, better predict which treatment will work best, and possibly even reduce the need for invasive biopsies, which are expensive, time-consuming and uncomfortable for the patient.
AI takes advantage of machines’ capacity to learn by using dynamic algorithms to enable new data to be assimilated on top of existing data to continuously refine knowledge. Radiomics is a growing field of medical study, in which greater amounts of information are extracted from medical images and from the patient’s electronic medical record, with the potential to uncover deeper layers of disease characteristics.
Used together, AI and radiomics aid in personalized cancer treatment by providing more information about an individual patient’s tumours, thereby helping oncologists to develop a customized course of treatment with a much better chance of success.
“In preliminary studies, the algorithms that we are formulating in radiomics allow us to elucidate molecular and other important features of the tumour with great precision,” says Dr. Forghani, a JGH radiologist and clinician-scientist.
“This provides us with information beyond what the most skilled physician can accomplish with the naked eye. We believe these models may enable us to better predict what is the most appropriate and least invasive treatment for an individual patient.”
Dr. Forghani says AI is also very exciting, because “it is helping us deal with a significant problem: physicians are finding it increasingly difficult, if not impossible, to integrate the large amount of information in a patient’s medical chart and scans in a way that is tailored to the individual patient’s care.
“Machine learning offers an opportunity to compare information that is derived from one patient’s tumour with all of the data that has been accumulated from a vast array of tumours, in order for us to predict how a comparable tumour is likely to progress.”
Eventually, he says, the JGH should serve as a hub for radiomics and other medical applications of machine intelligence, in collaboration with other academic institutions inside and outside Quebec.
Dr. Forghani adds that while the current focus of this program is on cancer diagnosis, many of the AI tools that are developed in this process have potential for broader applications—extending to non-oncological diagnostic and therapeutic prediction models, quality, safety and the cost-effective use of healthcare resources, in which the JGH can also play a pioneering role.