finn.behrendt(at)tuhh.de   +49 (0)40 42878 3547      E.3087

Deep Learning for Medical Analysis

For our research, we consider modalities such as optical coherence tomography, endoscopic images, MRI images etc. However, the large data streams and the often non-linear behavior of the observed systems demands advanced image processing algorithms. We therefore couple our approaches with state-of-the-art deep learning methods. Further, due to the scarcity of labelled data in medical domain, we try learning techniques such as self-supervision, contrastive learning etc on unlabelled data. Based on these, we can create custom deep learning pipelines to overcome the task specific challenges.

Please contact me directly if you are interested in any of the above topics.

  sarah.grube(at)tuhh.de   +49 (0)40 42878 4639     E.1051

Mobile Medical Robot

We offer several topics in the field of mobile medical robotics with the goal of autonomously examining patients, e.g. using different imaging modalities. Possible topics could be:

• Navigation of a mobile robot (path planning, SLAM,...)
• Robot positioning
• Autonomous patient scans

Ultrasound Imaging

Ultrasound is a common and important method in the clinics for different medical interventions. We have several ultrasound probes in our lab which we use for medical imaging and tracking. For this purpose, we use both conventional methods as well as deep learning. Possible topics for theses would be for example:

• Speckle tracking
• Ultrasound shear wave elastography
• Ultrasound image reconstruction
• Tracking biopsy needles during insertion

More specific topics regarding these issues can be found on our StudIp page. Please contact me directly if you are interested in one of these topics.

  suresh.guttikonda(at)tuhh.de   +49 (0)40 42878 2350     E.1052

Deep Learning for tbd ...

Please contact me directly for available thesis topics.

   sarah.latus(at)tuhh.de   +49 (0)40 42878 3389    E.3083

Image Guidance

Various imaging modalities can be applied to guide medical interventions, e.g., to optimize the instrument positions. We investigate methods to register multi-modal image data, including accurate temporal synchronisation and spatial calibration. 

Elastography

A correlation between soft tissue stiffness and the presence of tissue abnormalities has been demonstrated. Hence, a quantification of soft tissue elasticities is of great value in medicine. For this purpose, we study different methods for shear wave elastography based on ultrasound or optical coherence tomography. We analyze the variation in wave excitation and sampling and its influence on the estimated elastographic parameters.

Miniaturized Imaging Probes

The interopertive assessment of soft tissue properties is important to optimize the fine navigation of instruments and therby the treatment results. We aim to sense soft tissue characteristics from miniaturized imaging probes. We embedd optical fibers and with different imaging optics in medical instruments and investigate the quantification of morphological and mechanic tissue properties. 

  lennart.maack(at)tuhh.de   +49 (0)40 42878 3547      E.3087

Surgical Computer Vision

Computer vision, a field of computer science nowadays often based on deep learning methods, can be used to process and analyse surgical video/image data available due to the widespread use of endoscopy/laparoscopy including robotic surgery. In order to obtain valuable information for the surgeon, we use deep learning methods to analyse large amounts of image and video data. This includes the development of efficient methods to ensure real-time capabilities. Further, due to the scarcity of labelled data in medical domain, we try learning techniques such as self-supervision, contrastive learning etc on unlabelled data.

Deep Learning for Medical Image Analysis

For our research, we consider modalities such as endoscopic images, MRI and CT images, as well as Histopathological images. However, the large data streams and the often non-linear behavior of the observed systems demands advanced image processing algorithms. We therefore couple our approaches with state-of-the-art deep learning methods. Further, due to the scarcity of labelled data in medical domain, we try learning techniques such as self-supervision, contrastive learning etc on unlabelled data. Based on these, we can create custom deep learning pipelines to overcome the task specific challenges.

Please contact me directly if you are interested in any of the above topics.

   robin.mieling(at)tuhh.de   +49 (0)40 42878 3828      E.3088

Tissue-Needle Interactions

Feedback on local forces and tissue characteristics is limited in minimally invasive procedures. We therefore aim to develop novel methods for providing sensory feedback on tissue-instrument interactions to improve cancer detection. We currently employ optical coherence tomography systems for the image-based measurement of tissue elasticity and needle tip forces.

Robotic Needle Insertions

The acquisiton of tissue samples via percutaneous biopsy is a cornerstone of the diagnosis of cancer. Biopsy is conducted for sampling tissue in numerous target organs such as the prostate or the liver. We investigate how robot-assisted biopsy can further improve accuracy, reliability and safety of the sampling procedure.

Deep Learning for Medical Image Analysis

or our research, we consider high-resolution and high-speed imaging from modalities such as optical coherence tomography and ultrasound. However, the large data streams and the often non-linear behavior of the observed systems demands advanced image processing algorithms. We therefore couple our approaches with state-of-the-art deep learning methods. We utilize robotic manipulators together with our imaging systems to acquire labeled datasets. Based on these, we can create custom deep learning pipelines to overcome the task specific challenges.

Please contact me directly if you are interested in any of the above topics.

  maximilian.neidhardt(at)tuhh.de   +49 (0)40 42878 4639      E.1051

OCE-Imaging

Optical Coherence Tomography is an optical imaging modality based on interferometric measurement of reflections of near-infrared light from different tissue depths. It offers a field of a view few millimeters but micrometer-scale spatial resolution.

The very high temporal resolution of modern OCT systems make it ideal for imaging shear waves which travel at high speed through tissue. The spreading of the wave can be directly linked to the elastic properties of a potential malignant tissue.

Ultrasound Imaging

Medical ultrasound has been long established in the clinical routine and is still today advancing. Our lab offers two research ultrasound machines which can be used in a variety of scenarios such as:

• Tracking approaches with a robot
• Online image segmentation
• Shear wave elastography imaging.
• Image processing (beamforming)

Please contact me if you are interested in one of the above topics. In general you should bring enthusiasm for experimental lab work, good programming skills (C++/Python/Matlab) and the ability to work independently.

  konrad.reuter(at)tuhh.de   +49 (0)40 42878 3547      E.3087

Deep Learning for Medical Navigation

Please contact me directly if you are interested in any of the above topics.

  adrian.rudloff(at)tuhh.de   +49 (0)40 42878 3828      E.3088

Deep Learning for Medical Image Analysis

Please contact me directly for available thesis topics.

  schlaefer@tuhh.de   +49 (0)40 42878 3050      E.3085

Generally, I supervise all theses offered through our institute and I will be actively involved in theses projects. For more information, please see the topics offered through the research assistants.

However, there are number of toy projects I foster myself. Often these are longer shot problems that may have some impact in the future. For example, if you are interested in mobile robotics, please contact me. Also, I am happy to discuss your ideas if they are related to our research.

  johanna.sprenger(at)tuhh.de   +49 (0)40 42878 4639      E.1051

Image Guidance and Navigation

Image guidance, motion compensation and navigation are important in medical scenarios to support the surgeons and keep the patient safe. Different imaging modalities can be considered for these tasks, e.g. optical coherence tomography, ultrasound or camera systems.

Deep Learning in Medical Imaging

Medical imaging technologies deliver valuable information about the patients health. We have different medical systems in our lab and apply conventional image processing and deep learning methods to the acquired data. Different research projects are for example:

• Robotized surface scans and volume stitching with optical coherence tomography
• Robotized scanning of phantoms with ultrasound
• Needle placement and tracking with ultrasound guidance