Project summary B3CARE
Lung Cancer, Chronic Obstructive Pulmonary Disease (COPD) and Cardiovascular Disease (CVD), the so-called Big-3 (B3), are expected to cause most deaths by 2050. Early detection and prevention are crucial to lower the disease burden. Innovative low-dose computed tomography (CT) allows simultaneous, integrated assessment of early imaging biomarkers of lung cancer, COPD and CVD. Major advantages of integrated B3 screening are expected due to shared risk factors, B3 interdependence and health economic yield comparison to single disease screening. However, a major impediment before establishing a B3 screening program is the lack of validated and standardized B3 imaging biomarkers. The aim of this project is to advance the technology readiness level (TRL) of integrated B3 screening with at least 2 levels. To reach this objective, B3CARE will develop a large, high-quality imaging data biobank to provide biomarker reference values and validate B3 biomarkers using novel image analysis software and novel machine learning approaches. This will accelerate software development for CT-imaging data post-processing. Also, the expected substantial health economics potential of combined B3 imaging biomarkers for personalized health strategies is evaluated. B3CARE will thus provide an invaluable resource for the accelerated development and implementation of B3 imaging biomarkers and computer aided decision support. For more information about the entire B3CARE project, visit: https://www.b3care.nl/.
UT RAM Contribution
The B3CARE project is divided into 7 work packages (WPs), of which each WP has a different focus. The RAM contribution of this project is associated with WP5: Training & Education. The goal of this WP is to develop an educational program for (technical) physicians to train them to exploit computer assistance based on machine learning/post-processing software. It is one of the instruments for the realization of a widespread implementation of B3 screening. To accomplish this goal, several tasks are considered. A simulator of patient cases will be developed, that is able to mimic data of different patient cases with adjustable degree of complexity. This simulator can be used in a training and assessment program for radiological residents and technical physicians. In turn this simulator will be used to create a Computer Assisted Learning (CAL) system, also containing a library of trained machine learning algorithms. To give radiologists more insight in machine learning and its applicability to medical image data, a post-academic course ‘Machine Learning for Medical Applications’ is created.