AI from Chalmers helps doctors find heart and blood arteries disease
AI research conducted at Chalmers together with the SCAPIS project at the Sahlgrenska University Hospital was recently featured in the Swedish news. Based on images from the more than 30,000 participants in the SCAPIS project, Chalmers researchers are currently teaching computers to outline coronary arteries of the heart and to identify the areas with plaque and fat that could lead to future myocardial infarctions.
This is the first healthcare project at the national AI center, AI Innovation of Sweden, that was recently opened at Lindholmen Science Park in Gothenburg. The project is a close collaboration between the Sahlgrenska Hospital, Sahlgrenska Academy and Chalmers University of Technology.
One of the researchers involved in the project is Jennifer Alvén, PhD student in the Computer Vision and Medical Image Analysis research group led by professor Fredrik Kahl, at the Department of Electrical Engineering at Chalmers University of Technology in Gothenburg. Jennifer conducts research in the field of medical image analysis. Her research focus is machine learning, as well as explicit and implicit shape models, for classification and segmentation of medical 2D and 3D images. Examples of current applications are PET image registration, pericardium segmentation, heart ultrasound classification and coronary artery segmentation and classification.
– The images are cardiac CTA (computed tomography angiography) of people who participate in the SCAPIS study, says Jennifer Alvén. An experienced cardiologist manually outlines the coronary arteries using a digital drawing board. These manual delineations are thereafter used, by me, to train the computer to recognize the vessels automatically by means of deep learning. It will probably take hundreds or thousands of manually delineated images before the computer can do the job. In the long run, the automatic software will analyze all 30,000 SCAPIS images, a number to large to be feasible for manual assessment.
In the SCAPIS project, they also collect large amounts of other data than images, such as blood samples, ultrasound, lung functionality tests, and information on socio-economic conditions. This means that at a later state, one can expect that AI will be able to find completely new patterns and connections.
The Swedish CArdioPulmonary BioImage Study (SCAPIS) is led by professor Göran Bergström at the Sahlgrenska University Hospital and Sahlgrenska Academy at Gothenburg University. It was initiated as a major joint national effort in Sweden to reduce mortality and morbidity from cardiovascular disease (CVD), chronic-obstructive pulmonary disease (COPD) and related metabolic disorders, all of which are important issues for public health. Its main goal was to characterize, in terms of phenotype and environmental and socioeconomic influences, a Swedish cohort of 30,000 men and women aged 50–64 years. SCAPIS capitalizes on the latest developments in imaging. In addition, innovative use of large-scale genotyping and recent developments in metabolomics and proteomics will facilitate the identification of new biomarkers and mechanisms for disease.