MRI photo of a human head.
ASSIST. lev dolgachov

Swedish healthcare is facing major challenges in the coming years. One in three people in Sweden will get cancer at some point in their lives, and many of them will receive radiation therapy. Due to the resent pandemic, a large care debt has accumulated. In the coming years, Swedish healthcare must pay for this debt in addition to the care that will be performed as usual. This challenge can be addressed only through more efficient planning and treatment strategies.

In the ASSIST project, the main goal is to boost efficiency of healthcare, by taking advantage of the advances made in deep learning, wherein a computer is trained to perform various tasks. Radiation therapy is an effective treatment method for tumours, complementing surgery and chemotherapy. However, radiation therapy demands time-consuming preparations that involve acquisition of medical images, segmenting the tumour and risk organs, and developing a treatment plan for treating as much of the tumour as possible without harming healthy tissue. Deep learning can be used in all these steps, to shorten the time for planning, which leads to increased patient through out and shorter queues.

To determine the most effective treatment plan for tumour patients, there is pressing need for observations sensitive to small-scale changes within the brain. In the ASSIST project, we develop models and data analysis techniques for advanced magnetic resonance imaging (MRI) scans for delineating the tumour border accurately, thereby aiding the deep learning algorithms to be employed for treatment planning.

A general problem with deep learning for medical images is access to training data, which is complicated by the GDPR and ethical rules. In ASSIST, we develop methods for so-called ‘federated learning,’ wherein computers can be trained without medical images having to leave hospitals. We also develop methods to create realistic synthetic medical images, which can be shared freely because they do not belong to a specific patient.

More information on the project's ITEA4 page ►

Research Group


News about ASSIST