FEATURE: LIFE SCIENCE


Are machine vision and AI the beating heart of tomorrow’s medical care?


From monitoring the elderly, to blood imaging and in vivo microscopy, machine vision and machine learning are an increasingly powerful combination in healthcare, writes Sarah Fisher


H


ealthcare is one of the most exciting frontiers in AI development, with the potential


to transform our quality of life and the way we administer medical care. It’s also one of the most complex arenas for the integration of non-human data collection and decision-making. We hold our clinicians and providers to the highest ethical and professional standards, and rightly so. AI and computer vision tools that automate or share tasks with human care providers must meet, if not exceed, this level of excellence. So how are innovators at the cutting


edge of AI healthcare navigating the particular considerations of their specialism? And what new capabilities are they achieving as a result?


Balancing safety and privacy – machine vision in older people’s healthcare Professor Robert Fisher has been in the computer vision field for more than 40 years, with research spanning 3D scene understanding, model-based object recognition, range image analysis and parallel vision algorithms. “Essentially, it’s understanding the shapes of things in the real world,” he says. Fisher works with the Advanced Care


Research Centre, a multi-disciplinary project at the University of Edinburgh that focuses on improving quality of life for older people. His team has developed a form of visual monitoring that contributes to helping older people living alone to stay in their own homes safely, for longer,


“We can use the data from these RGBD cameras to map a skeleton structure and extract information about how it’s moving”


while ameliorating the obvious privacy concerns that come with cameras in the home. “There’s the assumption that you either


pass away or become so debilitated that you have to move into some sort of care facility, and nobody really wants to do that. So, we’re trying to help people live in their own homes for that little bit longer. It saves them money, and it’s probably better for their health, including mental health, to be in their own environment,” he says. Computer vision in this context assesses data from cameras in a person’s home and identifies anomalies in behaviour and movement. These might indicate a health issue, either chronic (as in mobility deteriorating over time) or acute (as in a fall or fainting spell). The camera can be placed in a hallway, where it collects data each time the person passes by, or opposite a dining table or a favourite chair, building up an understanding of what is normal for the individual and judging whether something has gone awry. An important part of this is the capability of the camera to perceive


18 IMAGING AND MACHINE VISION EUROPE DECEMBER 2023/JANUARY 2024


depth as well as shapes and colours. “For each pixel, as well as a red, green and blue value, there’s a depth value,” says Fisher. “We can use the data from these RGBD cameras to map a kind of skeleton structure of the human body and extract information about how it’s moving. “It tells us things like how fast they’re


walking and whether the left leg and the right leg move at different speeds, which could indicate some limping, obstruction, hip problems or knee problems. We can give that information to the person or their doctor or physio, so they know something has changed and how,” says Fisher. “Another way we can use it is to track improvement, for example after a hip replacement, to make sure healing and recovery is on course.”


Anonymous skeletons The ‘skeletonisation’ is one part of what makes the monitoring anonymous, another being the fact that the images aren’t monitored by human beings, but processed by an AI. “There’s a machine- learning-based programme that extracts the skeletons from the camera data,” explains Fisher. “It’s been trained on large datasets with lots of examples and lots of markup indicating where the skeletons might appear. It’s a package that seems to work well enough once it’s trained that you can use it on practically anyone. It doesn’t need any tuning.” The camera and machine learning AI


work together to build a picture of the person’s movement based on multiple


@imveurope | www.imveurope.com


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32