Unmet Need
SPECT and PET imaging are two modalities which are widely implemented for the diagnosis, research and management of many diseases. The quality of their images is depleted by noise and other factors, causing the images to suffer from low resolution and artifacts. In addition, current iterative reconstruction techniques are extremely time consuming. Therefore, there is a need for a fast and accurate technique for reconstructing these images.
Technology Overview
Researchers at Johns Hopkins University have developed a machine learning system for nuclear medicine image reconstruction. The system can reconstruct activity images from raw projection data directly and compensate for imaging-degrading physics factors. The system uses two independent staged channels, one to accept projection data, and one for map attenuation. The technology provides images with superior quality and quantification compared to conventional algorithms.
Stage of Development
The invention is currently in the development and validation stage.