General Scientific Contact: Rudi Deklerck
Medical Image Processing
Scientific Contact: Rudi Deklerck
In this area we are focusing on the following topics:
Automated image processing and analysis schemes for the extraction of bio-medical parameters in order to assess the presence or evolution of a disease, in quantitative terms with the aim to aid in diagnosis and prognosis (CAD).
Examples:
RX-Mammography: change detection both at the level of microcalcifications, mass and architectural distortions in mammographical images of screening studies, confronting recent images with the ones of a former study.
CT, µCT-SPECT: characterization of tumors/lesions based on perfusion studies and their follow-up in time after treatment.
Algorithmic approaches: robust registration, 4D/5D segmentation- and pattern recognition methods, probabilistic geometrical deformable models and atlases.
Content Based Search: correlate multi-modal data (i.e. image data combined with other patient data such as lab results, etc.).
-
Advanced visualization and compression of higher dimensional data sets.
Mathematical Problems in Imaging
Scientific Contact: Bart Truyen
Medical Imaging has become a major research theme in applied mathematics. The underlying inverse problems often are ill-posed, which means that the process of recovering their unknown parameters is very sensitive to errors in the measured response. Emphasis traditionally has been on the application of advanced concepts from functional analysis. It is not until more recently that the phenomenal growth in computer performance has spurred an equally dramatic advancement in the field of numerical analysis. Bridging the gap between the diverse fields of inverse problems and numerical analysis may hold the key to a better understanding of some of the implementation aspects highly relevant to the development of robust solution methods. It is exactly on the verge of these two domains that we seek to contribute.
A prominent role in this research is reserved for the study of Electrical Impedance Tomography (EIT) / Electrical Capacitance Tomography (ECT). Both EIT and ECT are notoriously ill-posed nonlinear inverse problems, described in a rich literature, but for which many pertinent theoretical and practical issues remain unresolved.
Another research theme is concerned with the numerical aspects of retrospective projective geometric correction in differential imaging, such as embodied in digital subtraction radiography (DSR).
Contributions are made on the level of new applications, where intra-oral imaging, caries diagnostics, dental implantology, and ambulatory heart failure monitoring have been identified as rewarding new medical application fields to deploy EIT/ECT technology. Contrast enhanced DSR instead has been studied as a new caries diagnostic technique with high potential.
Research is also targeted towards the development of innovative reconstruction algorithms. Major themes here are the introduction of advanced subspace concepts, sparsity aspects, and the development of structure exploiting solution algorithms based on the application of displacement rank concepts.
Research portfolio (projects & collaborations):
Research is driven bottom-up, inspired by theoretical concepts, and structured top-down by the application fields studied. Both research levels are bridged by common technologies.
e-Health
Scientific Contact: Bart Jansen
e-health research is currently focusing on analyzing human motion in a variety of domains focusing on frail and elderly people with the long term goal to automatically detect behavioral changes related to (or even preceding) changes in the medical condition of the subject being monitored in a continuous and unobtrusive manner.
Accelerometer based gait analysis.
Physical activity monitoring using wearable sensors. Quantification of energy expenditure and physical activity intensity.
3D camera based telemonitoring of elderly. Quantification of physical activity based on video images.
Automated assessment of fall risk. Irregularity in the walking pattern increases fall risk. The establishment of well defined cut-off values for a few parameters fails to identify the fallers among the broad spectrum of gait disorders.
Other classifiers, decision support systems and medical signal processing.
Research