MPBE Lunchtime Seminar: Dr M. Jorge Cardoso

Quantitative Neuroradiology: From big data to personalised medicine

Abstract:

In the last 30 years, medical image analysis has evolved from a fringe research topic to a mature science. Many algorithms and mathematical models have been proposed to diagnose, prognose and improve our understanding of pathology. However, due to the complexity of clinical workflows, lack of algorithmic robustness and the quality of clinical data, very few methods have had a direct impact on patient care. The Quantitative Neuroradiology Initiative (QNI), a partnership between the Centre for

Medical Image Computing at UCL and the Neuroradiological Academic Unit at NHNN, has took on the challenge to translate image analysis techniques from the research lab to the hospital.  QNI has developed a computational hardware and software infrastructure within the hospital walls that can process, learn and predict from clinical data at scale, while being respectful of current clinical workflows, algorithmic safety, quality assurance and information governance issues. This infrastructure has enabled a unique research programme where advanced multivariate predictive models learnt from hundreds of thousands of datasets (big-data with real world clinical quality scans and population-representative anatomical variability) are translated to individual patients to support diagnosis, prognosis and treatment. Tackling this research frontier requires the development of new engineering solutions to automated quality control, algorithmic robustness to uncommon phenotypes, data curation for large-scale learning, natural language processing for radiological report understanding, etc., and a new general purpose machine learning approach to image modelling that is designed for the clinical world.

Bio:

Jorge Cardoso is lecturer in quantitative neuroradiology between the Centre for Medical Image Computing at the Department of Medical Physics and Biomedical Engineering and the Neuroradiological Academic Unit at the National Hospital for Neurology and Neurosurgery. He has a PhD from UCL, a BSc and MSc in Biomedical Engineering at Minho University, Portugal. His expertise lies on the development of novel algorithms for imaging biomarker extraction. He has over 150 peer-reviewed publications, including more than 50 journal papers. He is the technical lead of the Quantitative Neuroradiology Initiative (QNI), which aims to extract from clinical neuro-images robust and accurate objective measures that support diagnosis, follow-up and treatment, and integrate them into neuroradiological workflow systems (PACS/RIS). He is also the main developer of NiftySeg, a popular open source package for image segmentation (niftyseg.sf.com) with more than 4000+ downloads, and a founder of brainminer ltd., an SBRI funded start-up that aims to bring imaging biomarkers to help the diagnosis, prognosis and management of neurodegenerative diseases to clinical practice.