Our research area focuses on the development and application of advanced machine learning techniques for the analysis, integration and exploitation of multi-source biomedical data in the context of personalised medicine.

Machine learning describes a set of computational and statistical methods that can make computers learn to perform tasks based on existing data. These include dimensionality reduction (e.g. PCA or kernel PCA), multivariate regression (e.g. PLS), classification approaches (e.g. SVM), and more advanced techniques (deep learning). 

The personalised medicine tasks that are of particular relevance include computer-aided diagnosis, risk and prognosis estimation, as well as treatment planning. The biomedical data of interest can be from different sources, including imaging, clinical, demographics, lifestyle, and environmental data.

Here are some examples of machine learning applications in our lab:

  • Cardiac statistical analysis by integrating imaging, lifestyle, and clinical measurements in conditional PCA models (Pereanez et al. Statistical Atlases and Computational Models of the Heart 2015).
  • Bone strength quantification by integrating shape, mineral density, microstructural and biomechanical information based on PLS (Lekadir et al. Annals of Biomedical Engineering 2016).
  • Cardiac automatic diagnosis using shape and textural radiomics in an SVM classifier (Cetin et al. Statistical Atlases and Computational Models of the Heart 2017).
  • Atherosclerotic plaque characterisation in carotid ultrasound using deep learning (Lekadir et al. IEEE Journal on Biomedical and Health Informatics 2017).
  • Alzheimer’s disease stratification through kernel methods (Zimmer et al. 2015 Computerized Medical Imaging and Graphics 2015).