- Deep Learning derived Input Function in dynamic PET
About the technology
With DLIF we present a non-invasive, automatically generated input-function for kinetic modelling in dynamic PET. Our deep learning models are trained end-to-end with dynamic PET images as input and ready-to-use input function as output. The model show very strong correlation and non-significant difference in influx (Ki) for myocardium and tumor derived with AIF and DLIF, respectively.
Value proposition
- No surgery requirements
- Compatible with any PET scanner and the most widely used tracers
- Provides results that are highly accurate with less bias and variation than state of the art
Areas of application
- Non-invasive input function for dynamic PET where blood sampling is not possible
- Preclinical or clinical applications
- Could be integrated into external software independent of PET scanner vendor
- As a part of the PET scanner-software to deliver the DLIF together with the PET images
Resources and partners
- The Norwegian Nuclear Medicine Consortium 180°N
- Tromsø Research Foundation
- Centre for Research-based Innovation (SFI) Visual Intelligence
- Proof of concept validated for small scale research in mice
Comparison to state of the art technology

Opportunities for Collaboration
We are currently looking for industry partners who finds interest in our technology. We are open for working together on the development of DLIF to accommodate your needs.
Collaboration partners




Contact information

Les om andre prosjekter:
Predicting relapse in UC patients
– A novel biomarker with the potential to improve the lives of millions of UC patients world-wide About the technology Ulcerative Colitis …
DLIF
– Deep Learning derived Input Function in dynamic PET About the technology With DLIF we present a non-invasive, automatically generated input-function for …
Mission – VTE
– MicroRNAs for risk assessment and treatment of venous thromboembolism The challenge Venous thromboembolism (VTE) is a highly prevalent disease, affecting around …