Alexander Jaus

Alexander Jaus

Postdoctoral Researcher in Computer Vision

Hello World!

I'm Alexander Jaus

Postdoctoral Researcher · Computer Vision & Medical Image Analysis · CV:HCI, KIT

I build deep-learning methods and scalable pipelines for 3D medical image analysis — segmentation, lesion detection, and volumetric quantification across CT, PET/CT, and MRI.

Medical Imaging Segmentation 3D Vision Dataset Construction at Scale Evaluation Metrics
Explore my research

I’m a postdoctoral researcher at the CV:HCI Lab at the Karlsruhe Institute of Technology (KIT). I completed my PhD in November 2025 as part of the Helmholtz Information & Data Science School for Health (HIDSS4Health) — a doctoral program jointly run by KIT, the German Cancer Research Center (DKFZ), and Heidelberg University — advised by Prof. Rainer Stiefelhagen and Prof. Dr. Dr. Jens Kleesiek.

I work at the intersection of computer vision and radiology, aiming to make medical image segmentation scalable and clinically meaningful — generating large anatomical datasets with minimal expert input, integrating anatomical priors into pathology modeling, and developing structure-aware evaluation metrics.

News

  • Jun 2026 “Good Enough? On the Impact of Label Quality in Large-Scale Medical Datasets” received an early acceptance at MICCAI 2026 — placing it among the top 9% of submissions. Read more →

Research focus

My work advances medical image segmentation at scale by minimizing radiologist involvement in dataset creation. I develop methods to generate large anatomical datasets through aggregation and semi-automated refinement. Recognizing that large datasets inevitably include annotation noise, I investigate how label quality affects model performance under varying scenarios, including pretraining and task-specific fine-tuning. A core focus is leveraging anatomical priors to improve pathology segmentation, based on the hypothesis that pathology can be modeled as deviations from normal anatomy — which I formalized through the Anatomy-Pathology Exchange (APEx) model. To assess segmentation quality beyond pixel-wise metrics, I develop component-level evaluation methods (e.g. CC-Metrics) that better reflect clinical relevance.

Selected Publications

indicates shared first authorship

Let’s connect

I’m always glad to talk through research ideas, potential collaborations, or anything in between — whether you’re a fellow researcher, a prospective student, or simply curious about the work. Don’t hesitate to reach out.

Get in touch