Improvements in Scatter Correction for Brain PET/MRI

3 February 2020

Bo Ma, Michaela Gaens, Liliana Caldeira, Julian Bert, Philipp Lohmann, Lutz Tellmann, Christoph Lerche, Jürgen Scheins, Elena Rota Kops, Hancong Xu, Mirjam Lenz, Uwe Pietrzyk, Nadim Jon Shah

In PET imaging, scattering occurs when the photons emitted by the radioactive source interact with electrons as they pass through any matter on their path towards the detectors. This results in photons being 'knocked off course', causing degradation in the attained image.

By correcting scatter, image quality and quantitative accuracy can be improved. The most accurate method of scatter correction is currently achieved by using Monte Carlo simulation (MCS). However, this method requires computational times that far exceed clinical feasibility. The method is more commonly seen in the clinic, single scatter simulation (SSS), has the disadvantage of limited robustness for dynamic measurements and for the measurement of large objects

In this paper, the MCS is developed using graphics processing unit acceleration. The accuracy of the method was validated with simulations and its performance was compared to that of the SSS method. It was shown that the novel method was able to generate accurate scatter estimations within a clinically relevant time frame and outperformed the SSS method in terms of recovery coefficient and correction accuracy in phantom measurements.

These findings demonstrate the applicability of this method in the image reconstruction of real patient data. relaxation times of water molecules are dependent on the microstructural environment and, as such, can provide vital information about tissue damage in various neurological pathologies, such as multiple sclerosis, epilepsy, psychotic disorders, dementia, and traumatic brain injury.

Original publication:

Scatter Correction Based on GPU-Accelerated Full Monte Carlo Simulation for Brain PET/MRI