392284 ISY Project: Predicting justifiable seizure diagnoses from real medical patient data (Pj) (SoSe 2023)

The detection and diagnosis of seizures is a notoriously complex problem in today's medicine. Differentiating epileptic seizures from dissociative psychological triggers or cardiovascular causes is only possible through careful anamnesis, conservative test result interpretation and detailed semiological analysis. The idea of this project is to tackle this medical problem and its inherent incompleteness and uncertainty with respect to the occurence of symptoms and causing diseases from a computer scientific perspective. We will provide a data set of real patients comprising their gold standard seizure diagnosis and corresponding reports about their complaints. Your challenge is to build and compare models that are able to extract the underlying medical patterns in the data, to predict a diagnosis given a new patient and to derive algorithmic justifications on why this diagnosis is the correct choice. The results of the project will be summarized in a written report whose length depends on the group size. Possible research questions for this project include:
- Which (family of) algorithms is suitable for seizure diagnosis on a low-sample training set like this?
- Which (family of) algorithms produces interpretable diagnostic decisions and how?

In case the proposal would not attract enough students for a team project, it can be adapted into an individual project or a project for two students (tandem project).

Requirements for participation, required level

Required skills:
Basic machine learning knowledge