Techn. Fakultät Willkommen am Institut für Informatik FAU-Logo

Mario Amrehn M. Sc.

Researcher in the Segmentation (SEG) group at the Pattern Recognition Lab of the Friedrich-Alexander-Universität Erlangen-Nürnberg

Towards an optimized workflow in interventional medicine.
Robust Seed Mask Generation for Interactive Image Segmentation
Mario Amrehn, Stefan Steidl, Markus Kowarschik, Andreas Maier
  • In interactive medical image segmentation, anatomical structures are extracted from reconstructed volumetric images. The first iterations of user interaction traditionally consist of drawing pictorial hints as an initial estimate of the object to extract. Only after this time consuming first phase, the efficient selective refinement of current segmentation results begins. Erroneously labeled seeds, especially near the border of the object, are challenging to detect and replace for a human and may substantially impact the overall segmentation quality. We propose an automatic seeding pipeline as well as a configuration based on saliency recognition, in order to skip the time-consuming initial interaction phase during segmentation. A median Dice score of 68.22% is reached before the first user interaction on the test data set with an error rate in seeding of only 0.088%.

    Articles in Conference Proceedings
    2017 IEEE Nuclear Science Symposium and Medical Imaging Conference Record (NSS/MIC) (IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC)), Atlanta, Georgia, USA, 21.10-28.10.2017, pp. tbd, 2017 (BiBTeX, Who cited this?)
UI-Net: Interactive Artificial Neural Networks for Iterative Image Segmentation Based on a User Model
Mario Amrehn, Sven Gaube, Mathias Unberath, Frank Schebesch, Tim Horz, Maddalena Strumia, Stefan Steidl, Markus Kowarschik, Andreas Maier
  • For complex segmentation tasks, fully automatic systems are inherently limited in their achievable accuracy for extracting relevant objects. Especially in cases where only few data sets need to be processed for a highly accurate result, semi-automatic segmentation techniques exhibit a clear benefit for the user. One area of application is medical image processing during an intervention for a single patient.

    We propose a learning-based cooperative segmentation approach which includes the computing entity as well as the user into the task. Our system builds upon a state-of-the-art fully convolutional artificial neural network (FCN) as well as a simple rule based active user model for training. During the segmentation process, a user of the trained system can iteratively add additional hints in form of pictorial scribbles as seed points into the FCN system to achieve an interactive and precise segmentation result.

    The segmentation quality of interactive FCNs is evaluated. Iterative FCN approaches can yield superior results compared to networks without the user input channel component, due to a consistent improvement in segmentation quality after each interaction.

    Articles in Conference Proceedings
    Amrehn, Mario; Gaube, Sven; Unberath, Mathias; Schebesch, Frank; Horz, Tim; Strumia, Maddalena; Steidl, Stefan; Kowarschik, Markus; Maier, Andreas
    EG VCBM 2017 (Eurographics Workshop on Visual Computing for Biology and Medicine), Bremen, Germany, 2017-09-08, pp. 143-147, 2017, ISBN 978-3-03868-036-9 (BiBTeX, Who cited this?)
Comparative Evaluation of Interactive Segmentation Approaches
Mario Amrehn, Jens Glasbrenner, Stefan Steidl, Andreas Maier
  • Image segmentation is a key technique in image processing with the goal to extract important objects from the image. This evaluation study focuses on the segmentation quality of three different interactive segmentation techniques, namely Region Growing, Watershed and the cellular automaton based GrowCut algorithm.

    Three different evaluation measures are computed to compare the segmentation quality of each algorithm: Rand Index, Mutual Information, and the Dice Coefficient. For the images in the publicly available ground truth data base utilized for the evaluation, the GrowCut method has a slight advantage over the other two.

    The presented results provide insight into the performance and the characteristics with respect to the image quality of each tested algorithm.

    Articles in Conference Proceedings
    Bildverarbeitung für die Medizin 2016 (Bildverarbeitung für die Medizin), Charité - Universitätsmedizin Berlin, 13.03.2016, pp. 68-73, 2016, ISBN 978-3-662-49465-3 (BiBTeX, Who cited this?)
Portability of TV-Regularized Reconstruction Parameters to Varying Data Sets
Mario Amrehn, Andreas Maier, Frank Dennerlein, Joachim Hornegger
Transcatheter Arterial Chemoembolization (TACE)
  • Outline

    This project aims at semi-automatic detection and segmentation of focal liver lesions in volumetric C-arm CT images. CT images are commonly used for diagnosis, treatment planning and follow-up examinations of liver tumors. The C-arm CT images are acquired during the curative intervention after intravenous application of a contrast agent.

    To detect and segment the target lesions simultaneously, a voxel classification approach is pursued. During pre-processing, the search space is confined by segmenting the liver first.

    Hepatic lesion segmentation

    Primary liver tumors are among the most frequent malignant tumors. More than 600,000 people worldwide die of hepatic cancer every year, according to the annual reports of the Opens external link in new windowAmerican Cancer Society. Especially colon cancer patients have a high risk of developing liver metastases at some point. However, so far computer assistance for physicians during diagnosis and surgical treatment of liver tumors is usually very basic.

    This project aims at overcoming this deficit by providing an semi-automatic segmentation method feasible to use during an intervention. Given a set of CT images, it provides guidance for an accurate segmentation of liver lesions. The tumors' outline information is used to identify extra-hepatic collateral vessels feeding the cancerous tissue with oxigenated blood. Additionally, with an exact segmentation available, tumors can be monitored over a longer period of time in order to allow an assessment of tumor growth or shrinkage and thus the success of the treatment.

    A major contribution of this work will be the highly accurate and fast semi-automatic segmentation of liver lesions. In particular, we want to focus on inhomogeneous. Current semi-automatic segmentation methods, which are often based on thresholding, usually benefit heavily from well-shaped, homogenous tumors, but fail when given irregular cases. Therefore, machine learning techniques will be applied to robustly detect and segment these tumors in images.

    Transcatheter Arterial Chemoembolization (TACE)

    TACE is a minimally Invasive treatment for liver cancer. During TACE, a catheter is inserted into the femoral artery and subsequently navigated to the patient's vessels feeding the tumor with oxygenated blood.

    In order to successfully occlude all nourishing extra-hepatic collateral vessels (high efficacy) while preserving the surrounding healthy tissue (low toxicity), an exact segmentation of the hepatic tissue is crucial.

    The aim of this project is to enable and support an ultra selective TACE approach during the intervention by providing a framework to interactively perform an accurate and fast lesion segmentation in the interventional environment.