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

THIS WEBPAGE IS DEPRECATED - please visit our new website

Patient Monitoring for Neurological Diseases Using Wearable Sensors

Patient Monitoring for Neurological Diseases Using Wearable Sensors

In medical collaboration with:
Prof. Dr. Soheyl Noachtar, Oberarzt, Neurologische Klinik, Klinikum Großhadern
PD Dr. Kai Bötzel, Neurologische Klinik, Klinikum Großhadern
Dr. med. Jan Rémi, Neurologische Klinik, Klinikum Großhadern

Scientific Director: Nassir Navab

Contact Person(s): Loren Schwarz

Keywords: Surgical Workflow, Sensing


Quantitative analysis of human motion plays an important role for diagnosis, treatment planning and monitoring of neurological disorders, such as epilepsy, multiple sclerosis or Parkinson's disease. Stationary motion analysis systems in clinical environments allow acquisition of various human motion parameters based on inertial sensors or cameras. However, such systems do not permit to analyze patient movements in everyday-life situations over extended periods of time. Existing sys-tems using portable inertial sensors typically extract coarse-grained movement information, e.g. over-all activity indices. In this project, we investigate machine learning-based methods that allow us to recognize multiple activities and to track the human full-body pose from wearable inertial sensor data. We propose to employ a prior motion model to constrain the tracking problem from inertial sensors. Machine learning techniques, such as manifold learning and non-linear regression, allow us to build a prior motion model for each patient. The main component of the motion model is a low-dimensional representation of feasible human poses for the set of considered activities. After training, our method is able to recognize individual activities and to track the full-body pose of a patient, given only inertial sensor data. We estimate the current state of a person (activity type and exact pose in low-dimensional representation) by means of a particle filter. Using the learned representation of feasible poses for tracking significantly simplifies the search for suitable poses, as compared to exhaustively exploring the space of full-body poses.


Figure 1: Training phase.
Figure 2: Monitoring phase.


Contact Person(s)

Loren Schwarz
Dr. Loren Schwarz

Working Group

Diana Mateus
Prof. Diana Mateus
Loren Schwarz
Dr. Loren Schwarz


Technische Universität München
Institut für Informatik / I16
Boltzmannstr. 3
85748 Garching bei München

Tel.: +49 89 289-17058
Fax: +49 89 289-17059
Visit our lab at Garching.

Ludwig-Maximilians-Universität München
Campus Grosshadern
Marchioninistrae 15
81377 München

Lab - Room: 4K U1 912
Tel.: +49 89 7095 4606
Visit our lab at Klinikum Grosshadern.

internal project page

Please contact Loren Schwarz for available student projects within this research project.

Edit | Attach | Refresh | Diffs | More | Revision r1.3 - 07 Jun 2010 - 11:23 - LorenSchwarz

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