QEII Times: The computer that learns: QEII researchers strive for digital doctor

see original story here The future is now at the QEII Health Sciences Centre as researchers take their first steps toward artificial intelligence — a computer program with the ability to review tremendous sums of data, learn from its clinical colleagues, and in this case, even find cancer. The Biomedical Translational Imaging Centre, or BIOTIC, is an interdisciplinary group dedicated to the advancement and implementation of medical imaging technologies which, put simply, allow physicians to see a patient’s insides in the absence of a scalpel. Examples of these are MRI, CT and X-ray, all of which are interpreted, sooner or later, by radiologists. Dr. Sharon Clarke is one such radiologist and an active member of the BIOTIC team. Her line of work requires the routine interpretation of many thousands of these medical images, any one of which could contain information crucial to a patient’s wellbeing, from the outline of a tumour to the telltale fracture of a broken bone. “When you are on call at 4 a.m. with 1,500 images to look through, it would be nice to have a ‘second pair of eyes’ in the form of computer assisted diagnosis, to improve efficiency.” A computer program capable of sifting through these numerous images and pointing clinicians in the right direction has been on her mind for some time now. With the recruitment of computer science students Peter Lakner and Peter Lee through Dalhousie University, as well as support from physicists Steve Patterson and Chris Bowen, and pathologists Dr. Jennifer Merrimen and Dr. Cheng Wang, Dr. Clarke has spent the last couple years converting concept into computer code. “Artificial intelligence in radiology has become the hottest topic around,” says Dr. Steven Beyea, a physicist and scientific lead of the BIOTIC team. In the course of radiology conferences and from the lips of industry leaders, Dr. Beyea has heard the interest in, and inevitability of, these game-changing technologies. One of the major players in this forthcoming revolution is General Electric Healthcare, a subsidiary of General Electric (GE), which has invested heavily in the creation of exactly this sort of artificial intelligence. And with additional funds coming from the Atlantic Canada Opportunities Agency (ACOA), some of that GE investment has reached Dr. Clarke and the BIOTIC team. As she describes it, the program they’re building is able to review MRI scans of prostate cancer in which she’s already pointed out the tumour. In this way the program learns how to recognize the tumour, and with each passing lesson it becomes more and more capable of identifying prostate cancer on its own, theoretically without limit. “I really don’t think radiology as a field is under threat by this new technology,” says Dr. Clarke, well aware of the fear that computers will replace professionals, particularly among the professionals themselves. She says the artificial intelligence presently “in training” is similar to spell-check, supplying radiologists with a second set of eyes rather than replacing them outright. Much like spell-check, this program is susceptible to mistakes in its search for tumours and will always require human oversight, but in time it has incredible potential to support the efficiency and accuracy of radiological work. While it’s being put through its paces on prostate cancer, Dr. Clarke says there’s no reason it couldn’t be applied to every organ in the body, given time, expertise and continued investment. “I think this will be very helpful to radiologists,” she says, “integrated into our daily lives the same as any other medical tool.” For the time being, her digital pupil is still very much in its early stages and an end-date for technology this new is difficult to set, but Dr. Beyea has been working closely with GE Healthcare to plan out next steps, including considerations for testing, regulating and implementing. “It’s not a matter of if,” he says, “but when.”]]>