Computer Vision Biomarkers for Objective (Oculo-) Motor Assessment in PSP

Carla Palleis, LMU University Hospital

Progressive supranuclear palsy (PSP) causes problems with balance, walking, falls, eye movements, speech, and swallowing. At present, these symptoms are mainly assessed by a neurologist watching and scoring movements using a clinical rating scale called the PSP Rating Scale (PSPRS). While this scale is widely used, it depends on subjective judgment and may not capture small but meaningful changes. This makes it difficult to detect early disease progression or to evaluate whether a new treatment is working.

This project aims to make the assessment of PSP symptoms more objective and precise by using modern computer vision (CV) technology — a branch of artificial intelligence that can automatically analyze videos and measure movements. Using simple video recordings from a tablet, CV can detect and track how the head, mouth, body, arms, and eyes move, without using any markers or sensors. From these recordings, we can extract detailed information such as walking speed, stride length, or eye movement velocity.

We will study 40 people with PSP who have confirmed brain changes on a special tau-PET scan and compare them with 20 healthy individuals. Each participant will be recorded during routine clinical visits at the start of the study and again after 6 and 12 months. Using these data, we will train computer models to automatically score symptoms and track changes over time.

Our goal is to develop digital tools that can measure PSP symptoms more sensitively and reliably than human observation alone. CV-based, fully automated scoring of oculomotor and motor problems in PSP has the potential to make high-quality PSP assessment more widely available. In addition, it could enable earlier detection of disease progression, be implemented in multi-center clinical trials, and ultimately support the development of improved treatments for PSP.