Parkinson’s disease (PD) is a common neurodegenerative disorder that affects millions of people around the world. Unfortunately, the disease can be present for many years before it is formally diagnosed. Currently, doctors look for motor symptoms to diagnose PD, but these symptoms appear later, after significant damage to the brain has already occurred, resulting in a degree of permanent disability. Early disease diagnosis is often inaccurate.
An accessible site that reflects changes occurring in the brain is the eye, specifically the back of the eye known as the retina. This three-year project will use retinal scans to develop an automated, artificial intelligence (AI) method for PD screening, improving health outcomes through quicker and more accessible screening at early stages of the disease before motor symptoms occur.
A recent study has used UK Biobank funds image data and AI to differentiate PD from those without PD. Our project, in addition to funds images will use other eye imaging data to assess features of retinal layers for enhanced screening accuracy. Moreover, comparison to other neurodegenerative disease will be done to increase specificity to PD detection.
The development of the AI-based method is analogous to a teacher-student scenario. The computer will be presented with scans of Parkinson’s disease, Alzheimer’s disease and healthy participants’ eyes and will learn to categorize these scans. Then, random scans will be shown to the computer to recognize the group based on its previous learning. Once sufficient performance is achieved, the method will be tested for clinical applicability based on feedback from clinical collaborators.
To develop the automated AI-based method, access to a large and varied dataset is required. A diverse and extensive dataset allows the computer to capture the individual characteristics that differ from person to person. If the method is based on a small number of participants or lacks varied demographic characteristics, it may not perform well in clinical settings.
There has been a constant increase of digital data, in many cases manual processing becomes impractical. AI has been widely used in healthcare, has shown to improve the efficiency of diagnoses and could be used in city and rural environments.The method developed in this project has the potential to be implemented as a clinical screening tool. This tool would allow for quick assessments without the need for specialized equipment or staff, enabling easy screening that could improve referral accuracy, enhance health outcomes, and save healthcare resources.