My UM
My UM
Search

UM Data Science Research Seminar with Clinical Data Science Maastricht

-

The UM Data Science Research Seminar Series consists of monthly sessions organized by the Institute of Data Science, in collaboration with another department, faculty, or institute at Maastricht University. These collaborations aim to bring together scientists from all over UM to discuss breakthroughs and research topics related to Data Science. The upcoming seminar will feature researchers from Clinical Data Science Maastricht.

All events are in-person and free of charge. We also offer participants a free lunch.

Schedule

 

LECTURE 1: 12:00 - 12:30

Speaker: Yanqi Huang

Subject: "The AI-Augmented Radiologist: My Adventure at the Crossroads of Medicine and AI"

Abstract: In this presentation, I will share my journey as both a practicing radiologist and a young researcher navigating the evolving landscape where medicine meets artificial intelligence. I will discuss how intelligent image analysis approaches developed by clinical data scientists—including radiomics and advanced AI tools—can help address key challenges in modern radiology practice. Rather than focusing solely on technical specifications and performance metrics, I aim to convey the hopes and expectations we hold as medical professionals for the future of AI in healthcare. This talk aims to offer perspectives on the potential impact of AI tools on patient care and the radiologist's workflow.
 

LECTURE 2: 12:30 - 13:00

Speaker: Petros Koutsouvelis

Subject: "Deep Learning Approaches for the Segmentation of Focal Cortical Dysplasia" 

Abstract: Deep learning-based segmentation has become a cornerstone in medical imaging, offering automation of labor-intensive workflows and the potential to detect subtle abnormalities. However, the clinical readiness of these methods varies widely across diagnostic tasks. Focal cortical dysplasia (FCD), a leading cause of drug-resistant epilepsy, relies on accurate lesion delineation and surgical intervention for achieving seizure freedom. Yet, FCD lesions are often subtle, variable in shape and location, and frequently missed during visual inspection. Automated segmentation of FCD is an active field of research, but remains far from clinical deployment due to limited annotated datasets and high inter-patient variability.  This presentation explores the potential of modern deep learning approaches—including foundation models—to improve the segmentation performance of FCD lesions, while shedding light on the broader limitations of applying AI to data-scarce, diagnostically challenging tasks.

Organizers


Clinical Data Science 

 

IDS

Also read