Robust Interpretation of Machine-Learning-Extracted Features from OCT Data
This project combines visual processing, large language models and information retrieval to answer questions and extract information from complex documents.
For diagnosis and prognosis of Multiple Sclerosis (MS), the use of Optical Coherence Tomography (OCT) combined with AI can provide vital information on disease development. MS Centrum Pelt, has collected OCT data on 988 patients, constituting one of the largest datasets in the world. The project has successfully leveraged Optical Coherence Tomography (OCT) data combined with AI to advance Multiple Sclerosis (MS) diagnosis and prognosis. Working with MS Centrum Pelt, we have overcome the initial data access challenges by establishing efficient data extraction protocols and implementing anonymization scripts. The AI model has been successfully trained on a dataset of 740 OCT scans, significantly improving upon the initial 53-patient sample. The value of AI models in monitoring MS disease through OCT scans has been preliminary demonstrated with results documented.
Contact: Simon Appeltans
Researchers: Dirk Valkenborg (PI), Axel-Jan Rousseau (AI Researcher)
Research groups: UHasselt – DSI