Module CS4332-KP06

Duration

1 Semester

Turnus of offer

each summer semester

Credit points

6

Course of studies, specific fields and terms:

• Master Medical Informatics 2019, compulsory, medical image processing

Classes and lectures:

• Model and AI-based image processing in medicine (lecture, 2 SWS)
• Model and AI-based image processing in medicine (exercise, 2 SWS)

Workload:

• 20 hours exam preparation
• 100 hours private studies and exercises
• 60 hours in-classroom work

Contents of teaching:

• Methods and algorithms for the analysis and visualization of medical images including current research activities in the field of medical image computing. The following methods and algorithms are explained:
• Fundamentals of neural networks in medical image processing
• Convolutional Neural Networks and Deep Learning in Medical Image Processing
• U-Nets for image segmentation
• Autoencoder and Generative Adversarial Networks in Medical Image Processing
• Data augmentation techniques
• Random Decision Forests for the segmentation of medical image data
• Statistical shape models: generation and application for image segmentation
• ROI-based segmentation and cluster analysis for the segmentation of multispectral image data
• Live wire segmentation
• Segmentation with active contour models and deformable models
• Non-linear image registration
• Atlas-based segmentation and multi-atlas segmentation using non-linear registration
• 3D Visualization techniques in medicine

Qualification-goals/Competencies:

• Students can classify and explain advanced methods for medical image analysis on the basis of their characteristics. They can select these methods based on a given specific application.
• They are able to explain advanced methods of cluster analysis and classification especially with Convolutional Neural Networks and Random Decision Forests and to characterize them by their properties.
• You can explain the conception of neural network architectures of U-Nets, GANs or auto-encoders in detail. They can explain in detail the conception of neural network architectures of U-Nets, GANs or auto-encoders.
• They know prerequisites, problems and limits as well as augmentation techniques for the use of neural networks in medical image processing.
• They know different approaches to model-based segmentation, can describe the different model assumptions made here and are able to explain the optimization strategies and algorithms used here.
• They are able to assess the properties of various non-linear image registration methods and to select and parametrize similarity measures and regularization terms for a specific registration problem.
• They are familiar with methods of multi-atlas segmentation and can explain and exemplify the properties of different label fusion approaches.
• They can differentiate between different medical visualization techniques, classify them according to their specific advantages and disadvantages, and select and apply them in a meaningful way depending on a specific application problem.
• They can practically work on and solve problems in medical image processing using neural networks.
• They master the problem-related selection and implementation of data augmentation techniques, suitable network topologies and training procedures as well as the evaluation of the results.

Grading through:

• Written or oral exam as announced by the examiner

Is requisite for:

• Seminar Model and AI-based image processing in medicine (CS4333-KP04)

Requires:

• Medical Image Computing (CS3310-KP09)

Responsible for this module:

• Prof. Dr. rer. nat. habil. Heinz Handels

Teacher:

• Institute of Medical Informatics
• Prof. Dr. rer. nat. habil. Heinz Handels

Literature:

• H. Handels : Medizinische Bildverarbeitung 2. Auflage, Vieweg u. Teubner 2009
• T. Lehmann : Handbuch der Medizinischen Informatik München: Hanser 2005
• M. Sonka, V. Hlavac, R. Boyle : Image Processing, Analysis and Machine Elsevier, 2007
• B. Preim, C. Botha : Visual Computing for Medicine 2nd Edition, Elsevier, 2013

Language:

• German and English skills required

Notes:

Last Updated:

14.09.2021