Retinal Staging via Bayesian Reverse Modeling of pRF Data

Vision loss is a common condition in older adults, significantly impacting independence and quality of life. Yet the standard clinical test for detecting blind spots in the field of vision still requires patients to press a button whenever they see a small flash of light. These tests are slow, tiring, and often unreliable, especially for elderly or unwell patients who may struggle to maintain attention. This project develops a fully objective method to detect and measure these areas of reduced or missing vision using functional brain imaging instead of relying on patient responses. When we look at a visual pattern, specific regions of the visual cortex become active in a predictable layout that reflects the structure of the retina. By showing a simple moving pattern during a short MRI scan, we can record this activity and determine which parts of the retina are sending normal signals and which are not, without requiring any action from the patient. The project is designed in three stages. First, we test the method in healthy volunteers by deliberately leaving specific parts of the visual pattern unstimulated. This creates artificial blind spots in the stimulus whose size and position are precisely known. This allows us to assess how accurately the method can recover these areas from brain activity alone. Second, we extend the method to identify more complex and irregular shapes of vision loss, as seen in real eye diseases. This includes cases with partially reduced rather than completely absent vision. Third, we apply the optimized approach to patients with retinal conditions such as macular degeneration or other retinal disorders. We compare the results with standard clinical tests to evaluate how well the new method reflects true visual function. All scans utilize very short measurement times and require only passive viewing, thereby reducing the burden on patients. The analysis uses a specific statistical approach, namely Bayesian statistics, that estimates the most likely pattern of vision loss directly from the data. This provides more quantitative estimates rather than subjective thresholding methods. This project is expected to provide the first fully automated and unbiased method to identify visual field defects directly from brain activity. It will improve clinical diagnosis, help monitor disease progression more reliably, and provide an objective tool for evaluating new treatments such as retinal gene therapy.