Neural networks increase the accuracy of monolithic PET detectors

(03-05-2021) Researchers Milan Decuyper and Mariele Stockhoff (Medisip research group) demonstrate that very good spatial resolutions can be achieved with neural networks, superior to nearest neighbour positioning.

Gamma ray detectors used in positron emission tomography (PET) scanners must combine high spatial, timing and energy resolution with excellent sensitivity. Current clinical PET scanners employ pixelated scintillation detectors with a spatial resolution limited by their pixel size. Another option is to use a monolithic crystal detector, read out by a photodetector array, which offers increased sensitivity and resolution. Already implemented in preclinical PET systems, such monolithic detectors may soon also appear in clinical scanners.
The neural network predicts the position of gamma events in 1 mm steps over a monolithic crystal detector (left). A quiver plot (right) shows spatial resolution on a colour scale and bias vectors as arrows.








The neural network predicts the position of gamma events in 1 mm steps over a monolithic crystal detector (left). A quiver plot (right) shows spatial resolution on a colour scale and bias vectors as arrows.


Monolithic PET detectors, however, bring their own challenges, such as a lengthy calibration setup and edge effects. Another key task when using a monolithic detector is to design an efficient and accurate gamma event positioning algorithm, with limited degradation in performance towards the edges of the crystal. To achieve this, researchers of Medisip (a research group of the department of Electronics and Information systems) have used artificial neural networks to create a high-resolution gamma positioning algorithm. 

“We chose to investigate neural networks as they can be trained to directly infer the continuous interaction position from the measured light distribution, based on example data,” explains first author Milan Decuyper. “They can learn to optimally process events near the edges and, once trained, positioning events is fast and parallelizable.”

Source: Physics World

More information

Their findings have been published in Physics and Engineering in Medicine & Biology: Artificial neural networks for positioning of gamma interactions in monolithic PET detectors.

Milan Decuyper, Mariele Stockhoff, Stefaan Vandenberghe and Roel Van Holen.

 

Contact

Medical Image and Signal Processing (MEDISIP) - Departement of Electronics and Information Systems 

Prof. Stefaan Vandenberghe

Campus Ghent University Hospital | Blok B - Entrance 36 | Corneel Heymanslaan 10 | 9000 Gent