The development of algorithms for the automatic and semiautomatic processing of medical image data became more and more important in recent years. On the one hand, this depends on improved image techniques, which allow increasingly finer virtual representations of the human body. On the other hand, this depends on improved computer hardware, which enables an algorithmic processing of data in gigabyte range in a reasonable time. The aim of this habilitation thesis is the development and evaluation of algorithms for medical image processing. Overall, the habilitation thesis consist of a number of publications, which are structured in three comprehensive topics: -Segmentation of medical image data on the basis of template-based algorithms -Experimental evaluation of open source segmentation algorithms under clinical conditions -Navigation to support intraoperative therapies The topic Segmentation of medical image data on the basis of template-based algorithms includes the development of several graph-based algorithms in 2D and 3D, where directed graphs a constructed via templates. This involves algorithms for the segmentation of vertebra in 2D and 3D. In 2D a rectangle and in 3D a cubic shaped template is used to construct the graph and calculate the segmentation result. Moreover, a graph-based segmentation of prostate central glands (PCG) is performed with a spherical shaped template to determine the border between the PCG and the surrounding organs. Based on the template-based algorithms an interactive segmentation algorithm has been developed and implemented that provides the segmentation result in real-time to the user. The algorithm uses different templates, however, needs only one user-defined seed point to perform the segmentation. In a further approach, the user can interactively refine a segmentation results by additional seed points. Thus, enabling also satisficing semi-automatic segmentations in difficult cases. The topic Experimental evaluation of open source segmentation algorithms under clinical conditions covers the intensive testing of freely available segmentation algorithms with patient data from the clinical routine. This includes the evaluation of the semi-automatic segmentation of brain tumors, like pituitary adenomas and glioblastomas with the freely available Open Source platform 3D Slicer. Thereby it could be shown how a pure manual slice-by-slice measurement of a tumor volume can be supported and speed-up in practice. Furthermore, the segmentation of language pathways in medical scans of brain tumor patients has been evaluated on different platforms. The topic Navigation to support intraoperative therapies includes the development of software modules to guide intraoperative interventions in different treatment stages (therapy planning, execution, monitoring). This includes the integration of the OpenIGTLink network protocol into the medical prototyping platform MeVisLab, which has been evaluated with an NDI navigation system. Furthermore, the conception and development of a medical software prototype to support an intraoperative gynecological brachytherapy has been introduced. The software prototype contained also a module for the enhanced visualization during a MR-guided interstitial gynecological brachytherapy, which enabled the registration of a gynecological brachytherapy device into the intraoperative dataset of the patient. The single modules lead to a comprehensive image-guided system for the gynecological brachytherapy in a multimodal operation suite. Thereby, the system covers the pre, intra and postoperative treatment stages during an interstitial gynecological brachytherapy.
|Translated title of the contribution||Segmentation of medical data and image-guided intraoperative navigation|
|Award date||14 Dec 2016|
|Publication status||Published - Dec 2016|
- Imaging analysis
- medical research
- Medizinische Informatik
- medical informatics