Back to search

FORNY20-FORNY2020

KVAL: Deep learning in dynamic Possitron Emission Tomography (PET) Imaging

Alternative title: Deep learning i dynamisk Possitron Emission Tomography (PET) bildebehandling

Awarded: NOK 0.50 mill.

Project Number:

337529

Project Period:

2022 - 2022

Funding received from:

Organisation:

Medical imaging is widely used to visualize tissue, tumors and biological processes both clinically in disease diagnosis and in animals for research purposes. There are different types of medical imaging where some focus on visualizing anatomical structures or tissues (ex CT, MRI ) while others visualizes functional changes or processes (ie SPECT, PET). The latter imaging technique uses different types of radioactive molecules, named tracers, for visualization. Depending on the tracer used, PET can be applied in various areas such as cardiovascular diseases, neurological diseases and oncology. As radioactivity can be quantified, this technique can also be used to map biological processes over time like a video. This can be used to determine which area of a tumor for instance is most active or map activity levels of different steps in a biological process. This is called dynamic PET and gives more detailed and in-depth knowledge compared to standard static PET mostly used today. Dynamic PET is dependent on kinetic calculation of the time-dependent concentration of the radioactive molecule in the blood, called the input function. The input function varies between different species and individuals. Today this is measured by arterial blood sampling. In humans this is however, laborious, time-consuming, and potentially painful, with risk for complications. Also, in small-animals, arterial blood sampling is hampered by limited blood volume and complex and terminal surgery for the animal. This has hindered the application of dynamic PET as a whole both in clinic and in research, and as such limited the potential knowledge gained from the research. This project aims to develop a non-invasive input function prediction model, based on cutting edge deep-learning methodology, representing a substantial leap forward in current state-of-the-art of dynamic PET.

With this project we were able to identify the most attractive and obtainable marketsegment for the technology that we will pursue in the further development. We were also able to establish a network of industrypartners both within PET hardware and software to advise on the commercial aspects of the project. Some of these can be potential future lisencees of the technology. Together with a positive Freedom to Operate analysis and support from commercial parties the project is positioned for further development and funding.

Funding scheme:

FORNY20-FORNY2020