Hello, my name is Zdravko and I am a research assistant / PhD student at the Karlsruhe Institute of Technology in the Computer Vision for Human-Interaction Lab (cv:hci). My project is part of the Helmholtz Information & Data Science School for Health (HIDSS4Health).

Please visit my Google Scholar page for the complete publication list.

Welcome to my research exploration into the interactive segmentation of medical images. Interactive segmentation involves training and validating a segmentation model with the aid of annotators who provide inputs like clicks, scribbles, or bounding boxes. This iterative human-in-the-loop process refines the model's predictions, especially valuable for annotating intricate 3D medical scans. Remarkably, what might typically take an hour for voxelwise annotation can be condensed to just ~3 minutes with strategic clicks and scribbles.

I work in close collaboration with the Institute for Artificial Intelligence in Medicine (IKIM) at the University Clinic Essen. There, we focus on bringing such models from research into the real medical practice.

If you as an interested student would like to work on tasks in the field of medical image segmentation and bring such models in clinical practice, shoot me an e-mail and come work with me on a Bachelor/Master Thesis or on an open HiWi position (all open positions are listed below).

Topics include but are not limited to:

• Semantic segmentation of medical images
• Interactive segmentation
• Integrating medical reports (text) as visual grounding for interactive segmentation (images)
• Domain Adaptation for medical imaging modalities (PET-->CT, CT-->MRI, etc.)

• Since February 2022: PhD Student at CV:HCI
• M.Sc. in Computer Science with profile Artificial Intelligence at Karlsruhe Institute of Technology: October 2019 - December 2021
• Student Research Assistant at Fraunhofer IOSB, Karlsruhe: April 2020 - November 2021
• B.Sc. in Computer Science at Karlsruhe Institute of Technology: October 2016 - October 2019

Master and Bachelor Students

• Matthias Hadlich, "Sliding Window-based Interactive Segmentation of Volumetric Medical Images", Master Thesis: 01.04.2023 - 30.09.2023
  • Parts of the results were integrated into an MICCAI 2023 Challenge Submission (3rd place for Domain Generalization): [paper]
  • Parts of the results are published into an accepted paper at ISBI 2024 : [paper]
• Franziska Seiz, "Interactive Domain Adaptation from Computed Tomography to Magnetic Resonance Imaging", Master Thesis: 01.05.2024 - 04.11.2024
• Lena Heinemann, "Interactive Segmentation of Medical Images using the Segment Anything Model and Grounding DINO", Master Thesis: 01.12.2023 - 03.06.2024
• Isabella Mebus Kishi de Oliveira, "Children Alphabetization using Optical Character Recognition for Handwritten Text", Bachelor Thesis: 20.06.2024 - 21.10.2024

HiWi Students

• Hoang Nam To, HiWi Position: Large-scale Data Collection and Evaluation Protocols for Interactive Segmentation [pdf] - 01.02.2024 - 31.07.2024

• Lecturer, Deep Learning for Computer Vision II: Advanced Topics, WS 24/25
• Organizer of the  Practical Course Computer Vision for Human-Computer Interaction, SS 24
• Lecturer, Deep Learning for Computer Vision II: Advanced Topics, WS 23/24
• Organizer of the Practical Course Computer Vision for Human-Computer Interaction, SS 23
  • Results from one supervised project are published at MIDL 2024 : [paper] [code]
• Supervisor at the Seminar Computer Vision for Human-Computer Interaction, WS 22/23
• Supervisor at the Practical Course Computer Vision for Human-Computer Interaction, SS 22
  • Results from the supervised project are published at ISBI 2023 : [paper], [code]

• 1st position in the CVPR 2024: Segment Anything In Medical Images On Laptop (Task 2: Scribble) [paper]
• 3rd position in the AutoPET MICCAI 2023 Challenge for Domain Generalization (Team cv:hci) [paper]
• MICCAI 2023 Outstanding Reviewer Award (top 0.7%)
• 5th position in the AutoPET MICCAI 2022 Challenge (Team heiligerl) [paper]
  • In close collaboration with IKIM

• NCT Data Science Seminar at the German Cancer Research Center, Heidelberg, 10/2024 (link)