Parkinson’s disease is a slowly progressive neurological disorder in which
brain cells gradually die off. An accurate assessment of a patient’s condition
is essential for appropriate therapy. One test used to evaluate symptom severity
is digital spirography, where the subject is asked to draw an Archimedean
spiral. Any anomalies that appear during drawing (e.g., partially straight
segments, tightening/compression, jaggedness) reflect symptoms of parkinsonism.
In this thesis, we focus on detecting spiral compression in an explainable,
human-understandable way. We developed a transformer-based model.
As is often the case in medicine, we did not have access to a sufficient amount
of labeled data. Therefore, we generated a synthetic dataset and parameterized
the anomalies based on their interpretations. We evaluated the model
on a test set and via an interactive web interface developed for this purpose,
which enables spiral drawing and spiral-compression assessment using our
model.
|
