Hervé Lombaert: The Language of Shapes in Medical Imaging
Doctors must often work with poor quality images to establish their diagnosis. These noisy and overwhelming images, with barely perceptible contours, are difficult to interpret. For Hervé Lombaert, a professor in the Department of Software Engineering and Information Technologies, improving the quality of medical imaging involves artificial intelligence, computer vision and mathematics.
Advances in Medical Imaging Through Artificial Intelligence, Computer Vision and Mathematics
Professor Lombaert’s research is centred on artificial intelligence, computer vision, mathematics and medicine. His work focuses on analyzing complex shapes to detect biological anomalies, applying statistical learning and computer vision to better understand living beings, and analyzing medical images to model populations and illnesses.
Developing More Accurate Models for Reading the Brain
According to Professor Lombaert, the way we analyze shapes is not optimal, because we approach them based on extrinsic geometric information and we simplify, or even ignore, their intrinsic nature.
Analyzing brain surfaces is one example. Brains are composed of a multitude of folds of varied shapes that are geometrically very complex. However, we still approach the brain as a simple sphere, which greatly reduces the precision of data.
Professor Lombaert’s work consists of developing algorithms around the geometry of shapes and data to create more accurate digital models.
Integrating the Notion of Time into Heart Imaging
Professor Lombaert has made great strides in interpreting images of the heart and detecting anomalies by representing its structure through specific mathematical models. These models take time into account, in addition to the three usual dimensions, as the heart is constantly changing shape.
Spectral Graph Theory to Address Complex Shapes
Professor Lombaert is also interested in spectral graph theory. The need to address the tiniest details of shapes has led to the development of a new paradigm to establish statistics for complex biological shapes.
Artificial Intelligence to Analyze Medical Imagery
Despite all the machine-learning algorithm methods that have been developed to date, analyzing medical images remains a challenge. Professor Lombaert is developing statistical approaches to improve automatic medical image analysis by concentrating more specifically on working with the geometry of shapes and data.
Mapping Shapes to Detect Anomalies
Professor Lombaert is also working on maps of the average size of an organ and its variability in a given area. By knowing the standard information, an individual’s anomalies can be easily detected. His first atlas of human cardiac fibres has been published, a milestone in the field of heart modelling.