Leveraging Expert Knowledge for Medical Image Segmentation

This PhD and postdoctoral opportunity at the Czech Technical University, Faculty of Electrical Engineering, focuses on advancing medical image segmentation using deep learning and constraint optimization. The project addresses the challenge of limited expertly labeled data in medical imaging by developing novel algorithms that incorporate expert knowledge through constraints on invariance, topology, dimensions, position, count, shape, and other segmentation properties. The research will also tackle weak annotations, such as scribbles, and global properties like unbiasedness, aiming to improve segmentation quality and reliability. Key objectives include integrating deep learning with advanced constraint optimization techniques, such as penalty, barrier, and projection methods, the moving target method, and feasibility-guaranteeing reparameterization. These approaches will be combined with transformation-invariant network operators to create robust and efficient segmentation algorithms. The project is highly interdisciplinary, drawing on artificial intelligence, computer vision, machine learning, and biomedical engineering. Applicants should have strong programming, mathematical, and research skills, with prior experience in image processing and deep learning. The position is funded by the Czech Science Foundation, ensuring financial support for successful candidates. The research group is led by Prof J Kybic, whose academic profile and research interests can be explored at this link . Applications are accepted year-round, providing flexibility for prospective candidates. Interested individuals should prepare a CV and cover letter detailing their relevant experience and skills, and are encouraged to review the supervisor's research for alignment with their interests. For more information and to apply, visit the project page .