Student Projects

Candidate: 

Alexey Györi

Supervisor: 

Dr. Ekaterina Kutafina

From --- to 07/2018

 

 

Efficient planning of hospital bed usage is a necessary condition to minimize hospital costs. A precise forecast for up to several months can help to better plan the staffs’ holidays. A Nonlinear AutoRegressive eXogenous model (NARX) with Recurrent Neural Networks (RNNs) has been shown to produce good results in comparison with other published methods by a previous paper of our department [1]. This model combines historical occupancy data with external (exogenous) features for prediction. These features include public and school holidays as well as the weekday and do not include sensitive data. This project aims to build a web service to make this model accessible: Hospitals can upload their bed occupancy data and the server performs both, training and prediction with some of the hyperparameters being customizable. After the forecasting is finished, users are automatically informed of the results.

Candidate: 

Sascha Welten

Supervisor: 

Dr. Stephan Jonas

From 10/2017 to 06/2018

 

 

This report is an improvement for the preliminary work by Hankammer. The prior work proposed an Android based EEG visualisation framework using Visual Programming concepts called BrainLabP&V. Initial analyses revealed further potential improvements for the BrainLabP&V framework in the area of usability, flexibility, and extensibility. This work present three design changes applied to the given framework. Namely, we implement the Composite Pattern, the Factory Pattern, and we designed the application in the spirit of the Plug and Play approach. As the last part of the update, we refresh the GUI of the framework and provide new possibilities for interaction with the framework.

Candidate: 

Armin Mokhtarian

Supervisor: 

Dr. Stephan Jonas

From 03/2018 to 03/2018

 

 

Back pain is one of the most common medical complaints. Although caused many by different reasons, back injuries are preventable. The humans posture is a very important factor here. Since we do not see ourselves, we might notice a poor posture when it already causes back pain. This paper presents a Health Observer (HObs) system which is able to detect poor posture and notify the user through an iOS application. The application gives immediate feedback on the current position and furthermore allows to evaluate the posture through the day.

Candidate: 

Nico Boehr

Supervisor: 

Dr. Stephan Jonas

From 06/2017 to 03/2018

 

 

Traction Force Microscopy is a technique for tracking the translation of cells using imaging methods. To enable this, fluorescent hydrogel beads are embedded in an extracellular matrix on top of which the cells are seeded on. The given task is to determine the distribution of bead size, bead aspect ratio and whether the spatial distribution of beads is random given images created through Fluorescence Light Microscopy. A second data set at hand is composed of Secondary Electron Microscopy scans for which the same characteristics are to be analyzed. The yielded results are used to examine the suitability of different hydrogel bead stiffnesses for traction force microscopy. We apply binarization and morphological operators as image preprocessing in order to separate beads from their surroundings. The length of the long and short axis of each bead (and thus the aspect ratio) is obtained by fitting ellipses to the shapes resulted from the preprocessed images. Spatial distribution is examined using Ripley’s K and L functions with edge effect correction as presented by Ripley.

Candidate: 

Simon Oehrl

Supervisor: 

Dr. Stephan Jonas

From 11/2016 to 11/2017

 

 

The risk of getting skin cancer is a threat that can be easily reduced by limiting the exposure to solar radiation. However, most people do not notice immediately when they have been in the sun for too long. On the other hand, many people take photos of themselves for social media platforms multiple times a day. A method to prevent the sunburn from occurring would be to analyze the photos taken by a user and warning them if it detects the formation of a sunburn. This paper shows that the color of the skin with and without a sunburn differs significantly for different color models.

Candidate: 

Lisa Prinz

Supervisor: 

Dr. Stephan Jonas

From 11/2017 to 11/2017

 

 

3D microscopies allow the examination of tissues for biologic or medical applications, providing different perspectives and enabling spatial cognition. This is especially the case if the 3D data is used in a virtual environment. MicroVR is a headmounted display based virtual environment for 3D microscopies. This study report depicts and discusses several methods to improve the MicroVR application. The visualisation of the 3D microscopies was to display RGB data. To facilitate the file selection the names of the microscopies were added to the 3D user interface. The existing raycasting algorithm was accelerated and improved further by implementing normal detection, true to scale representation, adaptive step sizes, Phong illumination, and classification for the transparency. To enable a more immersive user experience the scene was enriched with furniture. Realtime is especially important and hard to achieve for headmounted display applications and was one of the main decision criteria in the development process. At the end of the study report the results are discussed and further improvements are proposed.

Candidate: 

Pascal Ackermann

Supervisor: 

Marko Jovanović,
Dr. Mihaela Rusu

From 01/2017 to 11/2017

 

 

Traction Force Microscopy is a technique for tracking the translation of cells using imaging methods. To enable this, fluorescent hydrogel beads are embedded in an extracellular matrix on top of which the cells are seeded on. The given task is to determine the distribution of bead size, bead aspect ratio and whether the spatial distribution of beads is random given images created through Fluorescence Light Microscopy. A second data set at hand is composed of Secondary Electron Microscopy scans for which the same characteristics are to be analyzed. The yielded results are used to examine the suitability of different hydrogel bead stiffnesses for traction force microscopy. We apply binarization and morphological operators as image preprocessing in order to separate beads from their surroundings. The length of the long and short axis of each bead (and thus the aspect ratio) is obtained by fitting ellipses to the shapes resulted from the preprocessed images. Spatial distribution is examined using Ripley’s K and L functions with edge effect correction as presented by Ripley.

Candidate: 

Dominik Stegemann

Supervisor: 

Dr. Stephan Jonas

From --- to 10/2017

 

 

The common way to diagnose Parkinson's Disease and monitor the progress of a treatment is the Unified Parkinson's Disease Rating Scale (UPDRS), a questionnaire consisting out of 42 questions. Although the rating scale is the most commonly used scale, certain problems come up with it. First, the examination takes time, roughly between ten and twentyfive minutes per session. Then, the scale uses subjective perception as rating basis, which causes multiple problem as the rating may differ between two doctors and sessions. To face these problems, the idea of a digital based diagnostics, compatible with UPDRS, is described in the following work. A mobile device such as the Myo armband is used to measure tremors and motions of the patient. The captured data is then classified via comparison to mean data. This approach allows a more detailed and comparable rating, while still being translatable to UPDRS. In this work, the new approach is exemplary shown for two tests from UPDRS, namely the Pronationsupination movement test and the action tremor.

Candidate: 

Marco Grochowski

Supervisor: 

Dr. Stephan Jonas

From 09/2016 to 09/2017

 

 

Candidate: 

Simon Siewert

Supervisor: 

Dr. Stephan Jonas

From 04/2017 to 09/2017

 

 

Keeping patient or personal user data in clinical studies or other med ical applications private and protecting the patients identity is, and always has been, a very important issue regarding medical information systems. Usually, pseudonymization is used in order to hide the user’s identity. However, in some cases and especially in research, data protection standards or policies do not per mit the gathering of pseudonymization patient related data, because it is still pos sible to associate the data with the corresponding user. In this paper I propose a publickey cryptography based approach to grant full anonymization at every point in time, while being able to extend the users’ data sets by new records. By signing the data records with a user related privatekey, one is able to group the records created by the same user without having the ability to identify the user who created the records. A set of cryptographic schemes like DSA, RSA, ECDSA and ECGOST are tested and their respective performance is evaluated. RSA and DSA yield good computational performance. Using a RSA with a key size of 2048 Bit takes only about half an hour to complete a scenario of 2000 users and 10 5 records. Thus, it is possible to efficiently achieve full anonymization using the proposed cryptographic approach.

Candidate: 

Adrian Wagner

Supervisor: 

Marko Jovanović

From 10/2016 to 06/2017

 

 

Education in the field of physiotherapy targets the motor skills of its students, as well as their interpretation skills of movement. Today, learning these skills can only be achieved in a small group setting comparable to a class room. Showing a training module to the students is followed by allocating time for each individual student to evaluate and aid in the execution. Scheduling the time between students results in small time slots to actually train physiotherapy. The project Media Didactics Meets Wearable Computing (MediWeCo) aims for the goal to change the way physiotherapy is taught, by enabling courses to be held digitally. Using wearable sensor technology in conjunction with the ubiquity of stationary and, specifically, mobile computers, such as mobile phones, courses can be made available outside of the group setting and evaluation can be automated. This work introduces a streaming service for audio and video accompanied with data from wearable sensors, namely the Myo armband from Thalmic Labs. Teach ing courses remotely becomes possible by equipping the teacher with a Myo and video streaming setup. Enabling the students to save streams for later playback al lows the students to practice the learning material at convenient times in a home setting. Building the streaming service upon technology and protocols introduced in HTML5 achieves high compatibility with modern web browsers, without the need for additional plugins. This enables the goal of addressing as many students as possible by providing an out of the box easy to use service.

Candidate: 

Sebastian Rabenhorst

Supervisor: 

Dr. Stephan Jonas

From 04/2016 to 05/2017

 

 

This student research project deals with the positioning and occupation detection of chairs using active reader passive tag RFID (Radiofrequency identification) technology. For this purpose two tools are implemented. The positioning tool locates the chairs by calculating the distances of each chair to all antennas and then uses multilateration to calculate the position. The second tool takes these positions as input and shows the occupation status of each chair in a simple web client.

Candidate: 

Samuel Schüppen

Supervisor: 

Dr. Stephan Jonas

From 08/2016 to 02/2017

 

 

Health Observer (HObs) is a project to measure and evaluate physiological data in work environments with the help of a wearable device. A smartphone application is used to gather sensor data. In this work, this smartphone application is developed. This includes the gathering of the sensor data produced by the wearable devices microphone and accelerometers. Additionally, a Tizen based smartwatch companion application, to periodically show a questionnaire to support the evaluation of the sensor data, is developed.

Candidate: 

Thomas Schemmer

Supervisor: 

Marko Jovanović

From 09/2016 to 09/2016

 

 

Reinforcing the education in medicine through visual or haptic feedback is becoming more and more important in the current schooling approaches. Especially if the experiments are being executed on animals and are therefore scarce or ethically difficult it is useful to enable more students the same experience. This project uses a microscope camera feed and streams it into the local network, where the whole class can watch with mobile devices or PCs. Students and operators are able to mark specific areas in realtime to highlight important features or ask directed questions. A video on demand (VoD) service is also included: on starting a stream the professor may choose to save the video on the host computer, all markings will be saved and replayed as well. The onetomany architecture utilizes web sockets to establish peertopeer connections, which reduce the server load; the novel protocol WebRTC guarantees efficient streaming while reducing the overhead.

Candidate: 

Mirko Kugelmeier

Supervisor: 

Dr. Stephan Jonas

From 10/2015 to 09/2016

 

 

Sleep studies are used in diagnostics of various sleep diseases. For these studies, patients are often required to sleep in special sleep labs or submit to elaborate preparation. In the context of mobile health, the department of medical informatics currently develops a mobile sleep lab that is both easy and comfortable to use, and can thus be used to monitor the sleep of a patient over a longer period of time. An integral part of sleep studies, such as Polysomnography, is the detection of rapid eye movements (REM). In search for a lowcost solution, this paper evaluates the use of a Myo device to detect REM. The Myo is a wearable wristband which includes sensors such as accelerometers, gyroscopes and an electromyograph (EMG). We show that data recorded using the Myo’s EMG is not suitable to detect REM sleep.

Candidate: 

Fabian Beckmann

Supervisor: 

Dr. Ekaterina Kutafina

From 04/2016 to 08/2016

 

 

In a hospital, efficient usage of available resources is critical. In this paper we specifically deal with operating rooms (ORs). Hospitals need to schedule their ORs such that they are optimally occupied. An easy way to do this would be to overbook it, such that there is always someone waiting in line in case a surgery is quicker than estimated. However, this most likely would mean that patients would have to be sent home without surgery. us, an improvement in the duration estimation of surgeries could benefit both patients and the hospital. An anonymous hospital provided us with historical data about surgeries performed there. ese include information about procedures performed, as well as some pseudonymized information about personnel, but no information about the patients. We want to find out whether a good estimation of surgery durations can be made based on this data, and to evaluate what would be the best method to do so.

Candidate: 

Martin Kühn

Supervisor: 

Dr. Stephan Jonas

From 10/2015 to 06/2016

 

 

The dispensing of correct medication is an issue in most hospitals, as often wrong medication in terms of wrong drugs and inaccurate doses are handed out to the patients. Most errors are due to human mistakes. In context of this issue and as a means of health care education, the serious game Dr. Fill has been developed by the mHealth research group. A remaining task of this project is to analyze the impact and effects of the game to the players within the scope of user studies. The latest version of the project is fully playable, nevertheless there still remain issues to fix and features to implement for capturing results of future user studies. Within this student research project it is aimed to resolve these issues and to implement a foundation for upcoming user studies.

Candidate: 

Christian Plewnia

Supervisor: 

Dr. Stephan Jonas

From 12/2015 to 04/2016

 

 

Small structures can be captured layer by layer using digital microscopes. Understanding the threedimensionality of the produced data can be challenging, since regular image viewing software displays only a single layer at a time. There are more sophisticated visualization methods (e.g., volume rendering), but they are usually still displayed on a regular twodimensional computer screen. This paper aims for making a step towards providing a more intuitive threedimensional perception by describing a concept for a virtual reality environment for the examination of microscopy image data. Additionally, a corresponding prototype implementation that uses the Oculus Rift headmounted display is presented.

Candidate: 

David Laukamp

Supervisor: 

Dr. Stephan Jonas

From 11/2014 to 04/2016

 

 

This thesis deals with the application of wearable computing devices, specifically the Myo1 by Thalmic, in the area of hygiene training for medical professionals working in hospitals. Hand hygiene is one of the main factors to combat the number of hospitalacquired infections (HAIs). In Germany, each year about 500,000 people suffer from HAIs, resulting in 10,00015,000 deaths. Most of them are considered to be avoidable by committing to higher standards in hospital hygiene. The spread of HAIs mostly occurs by contact of the hospital staff with multiple patients paired with insufficient disinfection of the hands. To ensure proper hand disinfection, all medical professionals are trained to follow strict hand washing procedures. Such procedures consist of multiple different hand movements which each are performed for a period of a few seconds. A challenge with teaching the correct hand washing technique is the continuous supervision of multiple students. This requires the repeated observation of each student performing the procedure. The use of wearable devices to algorithmically assess the quality of the hand washing could simplify this task. Additionally, this also allows the hospital staff to easily perform selfchecks to decide whether the technique is still performed correctly. This work aims to assess the possibilities and limitations of using the Thalmic Myo to identify the individual gestures which are performed during professional hand washing procedures. A machine learning approach is used to recognize the performed gestures based on the data provided by the Myo armband.

Candidate: 

Simeon Keller

Supervisor: 

Dr. Stephan Jonas

From 04/2015 to 01/2016

 

 

There are many possibilities during the process from the admission of a patient to the discharge for things to go wrong. While there are many unforeseeable problem sources, errors in treatment should be proactively reduced. Several studies on drug administration errors have been conducted, exposing multiple errors during the preparation and the administration stage, e.g., missing a contraindication for a medical drug, or not fully following drug preparation instructions. Administering an intravenous (IV) injection too fast is among the most common errors. Here the device of administration is of interest: one the one hand infusions and drug pumps can be used for longer taking drug applications. These devices require explicitly setting the administration speed. On the other hand the situation might make the usage of a syringe necessary. Here one has to distinguish how the drug is applied. In contrast to subcutaneous injections, where the application speed is not important, rapid IV injections can result in serious damage, e.g., leading to overdose symptoms. While knowing the appropriate speeds for the given medicine is inevitable, the administrating person also needs to realistically estimate the application speed. Especially in cases where the drug administration can stretch over several minutes a trained perception is needed. Improved training of nurses has been suggested and shown to reduce drug administration errors [10]. The focus of this student project is set on improving the training for the perception of the application speed during IV injections. With this goal in mind we developed the mobile application MyoSyringe to support this training. It observes the drug application and visualizes further information to support the user. The application is written for Android and utilizes the Myo device to track the users arm movements and muscle activity. In this student project we further want to investigate the usability of the data produced by the Myo device.

Candidate: 

Andreas Burgdorf

Supervisor: 

Dr. Stephan Jonas

From 06/2015 to 11/2015

 

 

The diagnosis of sleep disorders is a complex and sometimes expensive procedure. A possibility to support classical diagnosis methods and allow a longterm monitoring of patients are wearable devices which are equipped with sensors which allow to measure movements and vital parameters. The goal of this project is to evaluate if existing wearable devices are able to fulfill the task of sleep monitoring in practice and to produce results which can be used for a later sleep analysis. Therefore, a suitable device has to be found, which has at least an accelerometer and a heart rate sensor which provide raw data. Further a framework has to be developed which allows to track activities in daily and nightly life and prepares the fetched data for a later analysis.

Candidate: 

Matthias Urhahn

Supervisor: 

Dr. Stephan Jonas

From 03/2015 to 08/2015

 

 

Hospitalacquired infections (HAI) (e.g. multidrugresistant infections) are an issue that many hospitals today face. One of the most efficient ways to prevent such infections is correct and gapless execution of hand sanitization by hospital staff. This issue motivated the creation of a project that aims to improve teaching and possibly monitoring of correct hand sanitization. The base goal in either scenario requires recognition of specific gestures. The application described in this paper was designed to help create datasets of gestures on which the recognition can be based.

Candidate: 

Yi Xu

Supervisor: 

Dr. Stephan Jonas

From 02/2015 to 07/2015

 

 

Currently, an anesthesiologist has to read patient data on an anesthesia machine or on a big wall­mounted display when performing anesthesia in operating rooms. This action forces the doctors to shift their attention from the patient, which makes it difficult to get the real­time data and performing the anesthesia at the same time. In order to improve this situation, we demonstrate in this paper how to visualize patient data on Google Glass in a surgery room for anesthesiologists. This concept of using a wearable hands­free device to assist the doctors is illustrated in two parts. One part is the User Interface design, and the other part is how we synchronize the real­time patient data with other devices in the surgery room. This project is a cooperation of the mHealth group of the university hospital of RWTH Aachen University and the OR.NET project, which works on secure and dynamic integration and networking in operation rooms and hospitals.

Candidate: 

Siamak Mottaghian

Supervisor: 

Dr. Stephan Jonas

From 08/2014 to 04/2015

 

 

Fall is one of the most common accidents and for all elderly people an everyday danger. A malicious fall may result in serious consequences, such as bleeding, fractur, and damages to the central nerve system. The development of a system which can detect a fall has received much attention in recent years. And such that system was repeatedly developed. However, we are developing a system that detects a fall, not only by one Component but through the communication of a plurality of components with each other, such as accelerometerand pulssensor. This study aims to implement a fall detection system by using an Androidbased smortphone equipped with a triaxial gravity accelerometer and triaxial magnetometer to record acceleration and tilt signals.

Candidate: 

Ralf Bettermann

Supervisor: 

Dr. Stephan Jonas

From 12/2014 to 03/2015

 

 

In snow sports a fall can result in a serious accident. Furthermore, the arrival of rescuers is difficult on mountains since a helicopter is maybe needed. If such a serious accident happens the victim might not be able to recover from the fall by himself. Reasons for this are that he either is unconscious or the injuries are too fatal. The skiing crash of the German race driver Michael Schumacher shows that it is important for the rescuers to know how the accident happened to deal with the situation correctly. Therefore, the fall and the direction of the impact needs to be tracked and visualized. Other vital signs like the heart rate are also helpful for the rescuers. In the following we tackle related work in the area of crash and accident visualization. Crash simulation is very common in the car industry since they need to avoid to use real crash tests with hardware prototype cars to save money. The goal is to use multiple simulation runs before a real car is used in a crash test. Johansson et al.[7][1] extract crash related information from text reports of car accidents and visualize the accidents afterwards. This is done by creating trajectories for the movement of each car which participated in the accident and then simulating and visualizing the movement. But they do not show the directions and speed of the present forces, just the situation how the accident occurred. Kuschfeld et al.[8] present means to simulate car crashes. They visualize spots on the car with same physical force applied with the same color as iso contours. With this aid one can see if the damage of the car is tolerable related to the given force. In this work they use efficient texture mapping for coloring the iso contours so that it can be achieved in realtime.

Candidate: 

Aaron Krämer

Supervisor: 

Dr. Stephan Jonas

From 08/2014 to 02/2015

 

 

Many people get injured due to a fall. It would help the medic to know more about the course of events and the state of health on his arrival. Therefore, a fallrecognition application is developed running on Android. Sensors which are connected to the application gather information about the course of events and state of health of a person. So called output modules display these information on remote devices, if a fall is recognized and the person does not react. This should help rescue teams to create an appropriate diagnosis in the field. In this paper the methods and implementation of this application and how one can add own modules are described in detail.

Candidate: 

Isabelle Tülleners

Supervisor: 

Dr. Stephan Jonas

From 04/2016 to ---

 

 

It is hard to learn the daily processes in a hospital for medicine students. For example, the workflow during a s​urgery is important but difficult to understand. Therefore, a tangible elearning environment can be helpful. That means students can directly interact with their environment and get feedback through physical interaction. This is a new approach in contrast to ex cathedra teaching and online elearning. In a tangible environment, students can learn without risk and time pressure which device is useful in which situation and how devices have to be used. Such an environment can be an interactive surgery room which uses virtual reality or canvases for videos and so on. This student project builds the foundation of such an interactive room. It starts by focusing on the detection of objects and the presentation of their usage. Later on, this could be extended to a more interactive learning environment. Here, a prototype setup will be developed based on RFID sensors and possible problems and difficulties will be documented. Within the setup, it should be possible to place different objects on a reader and receive corresponding information through audio or video feedback. The detection of objects will be performed via RFID tags which will be attached to them and an RFID reader which will be connected to a Raspberry Pi. The different RFID tags have to be registered to the RFID reader to create the assignment between tags and objects. The software on the Raspberry Pi should play an audio or video file depending on the tag presented to the RFID reader. Furthermore, the software on the Raspberry Pi should provide an API such that the environment can easily be extended with additional devices. The API should have a service which maintains multiple listeners. The interface of the listeners should provide functions to propagate event ids when a new tag is registered, a tag is deregistered or a tag is detected. This is needed to inform the additional connected devices about the use of an RFID tag. When this setup consisting of RFID in combination with the Raspberry Pi is runnable, the expectation of this work is to clarify the practicability and usefulness of it. If the setup is not runnable in the end, the expectation is to state whether this approach is feasible at all and where certain problems are located.

Candidate: 

Leon Staab

Supervisor: 

Dr. Stephan Jonas

From 04/2016 to ---

 

 

Examination of tissue at the cellular level is important for advanced applications in medicine today. It is widely used to create diagnoses in the field of histology and cytology, for example. Another aspect is the detection of alterations in cells treated with different drugs. While it is intuitive to view 2Dimages on a regular computer screen, the same does not apply for 3 dimensional image data. Thus, a prototype virtual reality (VR) environment for headmounted displays has been developed by Christian Plewnia, enabling users to examine 3D microscopy image data in VR by moving the 3D image with a game controller. The current functionality is going to be improved and extended within the scope of this work. One of the main goals is to improve performance, eliminating nausea and headaches appearing for some users, caused by the currently low frame rate. Another important aspect is the development of a suitable user interface, accessible for users actively using the application. UIs in virtual reality should be projected directly into the 3Denvironment as opposed to 2DUIs, which can be drawn at a fixed position on the screen. Furthermore, a solution for artifacts appearing when viewing a 3Dvolume from the side, as described in Plewnias paper, is ought to be found. The goal is to achieve a higher frame rate (frames per second) and to comply with VR guidelines and best practices in respect to the UI. These improvements should result in a better overall experience of the VR application, bringing it one step further for productive use.

Candidate: 

Ike Kunze

Supervisor: 

Dr. Ekaterina Kutafina

From 09/2017 to ---

 

 

Electroencephalograms contain a lot of information about the health of the brain. It is difficult to analyze them which is why medical students are mostly trained by experienced physicians in a mentorstudent relationship. The presented application offers a new possibility in the form of a serious game allowing students to learn about electroencephalograms whenever and wherever they want.