[Researchers at the Department of Psychology at the University of Cambridge in a research computer lab]
[Researcher analyzing an MRI scan]
[Zoe Kourtzi, Professor of Experimental Psychology at University of Cambridge working at desk]
Zoe Kourtzi (interview): “For a long time, for about 30 years, we did not have any disease modifying treatments. Now things have changed, and we see new treatments coming into the market, and for now we know we have the potential to identify patients early and identify the right patients for these treatments. So, our work has really been focussing on early predictions, early diagnosis when we have the first symptoms, or even before symptoms, can we identify individuals that may develop dementia in the future.”
[Screen showing full body MRI]
[Radiologist]
[Screen showing multi-organ MRI scan in process]
[Unidentified patient in scanner]
[Researcher Liz Yuanxi Lee conducting a cognitive test on Delshad Vaghari, asking questions and taking notes of Delshad’s response]
Liz Yuanxi Lee (interview): “The advantage of AI is they can detect details or relations between parameters or features that human beings cannot easily learn. For example, the images we’ll look at so there might be some tiny millimetre resolution changes we cannot see with our naked eyes.”
[Liz Yuanxi Lee]
[Liz Yuanxi Lee explaining using her computer screen MRI scans of four patients and their cognitive test results, and how AI algorithms can predict whether the patients could later have Alzheimer’s]
[Two MRI scans which the researchers’ AI algorithm compares against each other. The scan on the left-hand side shows a patient who is less likely to develop Alzheimer’s compared to the patient on the right-hand side who is]
[Delshad Vaghari, research associate at University of Cambridge, analyzing data and algorithms]
Delshad Vaghari (interview): “AI can do different things, but in this particular project we used AI for classification. When usually we have two groups or more than two groups. For example, controls and patients, apples and oranges, we can train models to learn what is an apple and what is an orange, who is a patient and who is a control healthy aging. So, the model can learn these two. We call these type of model supervised because we tell them what is what.”
[Zoe Kourtzi, Professor of Experimental Psychology at University of Cambridge working at desk]
Zoe Kourtzi (interview): “It learns to identify patterns in the data and in this case associated with specific patient diagnoses so it will know in the end whether a specific pattern that appears in the brain scans together with the cognitive scores means that a patient will be progressing or a patient will remain stable and then can make predictions about new patients that it hasn’t seen their data before. These predictions can be reliable because it has seen all these other data.”
[MRI brain scans on computer screen]
Zoe Kourtzi (interview): “I truly believe that having translated all of these really smart maths and algorithms into the space of neuroscience and understanding the brain and being able to validate the predictions and the models that these tools can make into the clinical practice is really truly something fantastic that I didn’t do by myself, but I think really that’s where science can make a big difference for social good.”
[Exteriors of the Department of Psychology at Downing College, University of Cambridge]
This script was provided by The Associated Press.