ThesesPage

Chair for Computer Aided Medical Procedures & Augmented Reality
Lehrstuhl für Informatikanwendungen in der Medizin & Augmented Reality

Diploma, Master and Bachelor Theses

Running Theses

Intraoperative Ophthalmic Scene Reconstruction for surgical AR (DA/MA/BA)

Ophthalmic microsurgical interventions require a high level of handling precision for the surgeon as even little errors can damage intricate structures inside the human eye. Furthermore, the microscopic top-down view through the microscope is a limiting factor for the surgeon's depth perception, motivating the use of augmented reality to enhance the surgical view. With the use of volumetric intraoperative Optical Coherence Tomograph (OCT), 3D imaging can be performed during the surgery, yielding an additional source of information to guide the surgeon. However, to display this additional data to the surgeon, proper integration into the surgical view has to be considered. One very intriguing way to integrate the information is to use a focus and context visualization method already used by Bichlmeier et al. [1], see Figure 1. To be able to provide perceptually plausible in-situ rendering of the acquired volumes as a focus-and-context visualization, first the 3D surface of the retinal surface needs to be reconstructed as a 3D surface. This can be achieved by applying stereo block matching using a calibrated stereo image pair. To properly position the iOCT overlay, the surface must then be aligned to the coordinate system of the OCT volume. This, however, cannot be done preoperatively due to the complex optical setup involving the patient's own eye. Therefore, the second part of this project is to devise an online alignment that can compensate for the (potentially changing) optical pathway affecting all imaging modalities. This online alignment can involve a simple surface alignment step as well as more complex algorithms to account for deformations due to the different optical pathways between the two modalities.
Supervisor:Jakob Weiss, Federico Tombari
Director:Prof. Nassir Navab
Student:Ekaterina Kanaeva
start-end:04/2018 -
Deep Learning for Tool Detection and Tracking in Microsurgery (Bachelor Thesis)

The aim of this project is the investigation of the state-of-the-art deep learning architectures and frameworks with the purpose of detection and tracking of instruments in retinal microsurgeries. An implementation of a deep learning based instrument detection workflow shall be provided at the end of the project.
Supervisor:Hasan Sarhan, Dr. Mehmet Yigitsoy
Director:Prof. Nassir Navab
Student:Luca Alessandro Dombetzki
start-end:01.04.2018 -
Addressing Artefact-Related Image Challenges In Automated Polyp Detection (Bachelor Thesis)

Deep learning techniques are becoming the state-of-the-art in automated polyp detection and the performance of these algorithms relies heavily on the size and quality of the training datasets. Specifically, endoscopic videoframes tend to be corrupted by various artefacts that impair their visibility and affect polyp detection rate. This thesis aims to tackle the issue of artefacts in endoscopic images by extracting knowledge from an artefact dataset that contains over 2,000 images with more than 17,000 annotated artefacts. Our first step was to participate in the ISBI 2019 Endoscopic Artefact Detection Challenge, where we implemented a RetinaNet? architecture and finished 3rd in the challenge. This object detection framework is then trained on polyp images while we explore different knowledge combination methods, such as Learning without Forgetting, to address the artefacts in these images. We hope that this improves polyp detection performance as well as further expands our comprehension of the effects of artefacts in automated detection in endoscopy.
Supervisor:Dr. Shadi Albarqouni, Roger Soberanis
Director:Prof. Dr. Nassir Navab
Student:Maxime Kayser
start-end: -
Evaluation of real-time dense reconstruction for robotic navigation (Master Thesis)

Supervisor:Nikolas Brasch
Director:Federico Tombari
Student:
start-end: -
Distributed SLAM - Jointly mapping 3D Geometry (DA/MA/BA)

Exploring an unknown scene and self-positioning within it, is a common and well-studied problem in Computer Vision which is known as SLAM (Simultaneous Localization and Mapping). Core fields of application are autonomous cooperative robotics and vehicles as well as tracking and detection systems in the medical domain. Traditional methods target a single system, equipped with image sensors, exploring the scene and building up a map for localization (e.g. a single robot or drone moving within an unknown environment). New approaches also incorporate information from other sensors such as IMUs, gyro or GPS. Another objective for the determination of the position is outside-in-tracking of an object via marker tracking with external sensors, thus providing the relative position of an object with respect to the tracking system. To overcome the line-of-sight problem of outside-in-tracking, and the singularity constraint of traditional SLAM methods, the project aims to develop a distributed SLAM approach. Multiple systems (referred to as sensor nodes hereafter), equipped with an image sensor, contribute to a common map of the scene for localization, while being also tracked by outside-in-tracking for accuracy. Thus, accuracy and applicability can be elevated with a distributed SLAM approach, combining the information of multiple sensor nodes and an external tracking system. Furthermore, the necessity of complicated and error prone calibration processes for individual systems within one application scenario can be avoided. The objective is to develop a generative distributed SLAM approach for challenging scenes and applications. Features like loop detection and closing, pose graph optimization, re-localization and mapping should be extended to a distributed approach, also enabling scalability.
Supervisor:Patrick Ruhkamp, Benjamin Busam
Director:Prof. Dr. Nassir Navab
Student:Joe Bedard
start-end: -
Privacy-Preserving Federated Learning in Medical Imaging (Project)

Supervisor:Dr. Shadi Albarqouni
Director:Prof. Dr. Nassir Navab
Student:Wasiq Kasam Rumaney
start-end: -
Few Shot Segmentation in Medical Imaging (Master Thesis)

Semantic segmentation does a pixel-wise classification to assign a class or background to each pixel of an image. This problem requires a very large data set of pixel level annotations, which is often unavailable or very costly to create. The aim of this project is to build a state of the art low shot deep learning technique for medical images, which can from few dense or sparse annotated medical image labels derive semantic segmentation of a new previously unseen class.
Supervisor:Dr. Shadi Albarqouni, Ari Tran
Director:Prof. Dr. Nassir Navab
Student:Abhijeet Parida
start-end: -
Human-like Medical Diagnosis Deep Network by Out-of-Distribution Sample Detection (Master Thesis)

Recently, deep learning has great success in various applications such as image recognition, object detection, and medical applications, etc. However, conventional deep neural networks with the softmax classifier are known to produce highly overconfident prediction even for abnormal samples. In other words, these models have no awareness of what they can and cannot do. This is the main difference between AI and human. Most humans can do certain tasks with good performance but for uncertain tasks, they say ‘I don’t know how to that. In this project, a new deep neural network, named human-like medical diagnosis network will be developed to solve abovementioned limitations in particular for safety-critic medical diagnosis tasks. The human-like medical diagnosis network assesses the difficulty of the query sample and rejects to answer on those cases. By rejecting to answer on difficult cases, it is possible to aware of the limitation and guarantees good performance on the cases it operates on. The central problem in the human-like medical diagnosis network is how to access the difficulty of the test sample. Measuring the uncertainty or difficulty of the prediction still remains a challenging task. In this project, we will develop a solution for this problem with the generative model for defining a new confidence score in the medical diagnosis applications.
Supervisor:Dr. Seong Tae Kim
Director:Prof. Dr. Nassir Navab
Student:
start-end: -
Incremental Learning For Robotic Grasping (Master Thesis)

Supervisor:Fabian Manhardt
Director:Federico Tombari
Student:Pengyuan Wang
start-end: -
Rethinking Deep Learning based Monocular Depth Prediction (Master Thesis)

We are looking for a motivated student who wants to work in the topic of monocular depth prediction using deep learning. Predicting depth from a single color image is a challenging and under-constrained task, and as such, active research is happening that incorporates CNNs. Recent works typically do not enforce an understanding for the objects in the scene. In contrast, our goal is to rethink depth prediction and create an object-aware model, which might lead to more accurate depth.
Supervisor:Helisa Dhamo
Director:Federico Tombari
Student:
start-end: -
Learn to learn: Which data we have to annotate first in medical applications? (Master Thesis)

Although the semi-supervised or unsupervised learning has been developed recently, the performance of them is still bound to the performance of fully-supervised learning. However, the cost of the annotation is extremely high in medical applications. It requires medical specialists (radiologists or pathologists) required to annotate the data. For those reasons, it is almost impossible to annotate all available dataset and sometimes, the only a subset of a dataset is possible to be selected for annotation due to the limited budget. Active learning is the research field which tries to deal with this problem [1-5]. Previous studies have been conducted in mainly three approaches: an uncertainty-based approach, a diversity-based approach, and expected model change [3]. These studies have been verified that active learning has the potential to reduce annotation cost. In this project, we aim to propose a novel active learning method which learns a simple uncertainty calculator to select more informative data to learn the current deep neural networks in medical applications.
Supervisor:Dr. Seong Tae Kim
Director:Prof. Dr. Nassir Navab
Student:
start-end: -
Photorealistic Rendering of Training Data for Object Detection and Pose Estimation with a Physics Engine (Bachelor Thesis)

3D Object Detection is essential for many tasks such a Robotic Manipulation or Augmented Reality. Nevertheless, recording appropriate real training data is difficult and time consuming. Due to this, many approaches rely on using synthetic data to train a Convolutional Neural Network. However, those approaches often suffer from overfitting to the synthetic world and do not generalize well to unseen real scenes. There are many works that try to address this problem. In this work we try to follow , and intend to render photorealistic scenes in order to cope with this domain gap. Therefore, we will use a physics engine to generate physically plausible poses and use ray-tracing to render high-quality scenes.
Supervisor:Fabian Manhardt, Johanna Wald
Director:Federico Tombari
Student:Wessam Abdelbari Ali Frrag
start-end: -
Trajectory Validation using Deep Learning Methods (Master Thesis)

Supervisor:Federico Tombari
Director:Federico Tombari
Student:
start-end: -

Finished Theses



Edit | Attach | Refresh | Diffs | More | Revision r1.14 - 02 Jan 2019 - 11:39 - TobiasLasser

Lehrstuhl für Computer Aided Medical Procedures & Augmented Reality    rss.gif