Data Science project for the classification of psicosis spectrum disorders (PSD) using resting state neuroimaging data (fMRI scans). This project was developed for the Data Science competition of the faculty of computer and information science (FRI) from the university of Ljubljana (UL). The results of our research suggest that the problem of binary classification of disease status of a patient (whether a person has a disorder or not) is solvable using a GBC (Global Brain Connectivity) representation of the fMRI data. We were also able to build a latent space representation of the GBC data where healthy and unhealthy individuals are clearly differentiable. However, our results for the classification of the specific disorders of the patients remained modest with an accuracy of about 50%.
- final_report: latex code and figures used to create the final report of the project. In this report you can read details of all the findings and methodology of the project, as well as references to additional material.
- interim_report: latex code and figures used to create the interim report required in the project.
- journal: Our personal journals detailing what each of the members of the team did and the hours of dedication to each task.
- presentation: Set of slides we used to present our team to the jury of the competition.
- src: all the code used to produce our results. It consists of analytical jupyter notebooks and also some stand alone python code for the more intricate implementations.