Concept development of an on-chip PET system

Abstract

Abstract Background Organs-on-Chips (OOCs), microdevices mimicking in vivo organs, find growing applications in disease modeling and drug discovery. With the increasing number of uses comes a strong demand for imaging capabilities of OOCs as monitoring physiologic processes within OOCs is vital for the continuous improvement of this technology. Positron Emission Tomography (PET) would be ideal for OOC imaging, however, current PET systems are insufficient for this task due to their inadequate spatial resolution. In this work, we propose the concept of an On-Chip PET system capable of imaging OOCs and optimize its design using a Monte Carlo Simulation (MCS). Material and methods The proposed system consists of four detectors arranged around the OOC device. Each detector is made of two monolithic LYSO crystals and covered with Silicon photomultipliers (SiPMs) on multiple surfaces. We use a Convolutional Neural Network (CNN) trained with data from a MCS to predict the first gamma-ray interaction position inside the detector from the light patterns that are recorded by the SiPMs on the detector’s surfaces. Results The CNN achieves a mean average prediction error of 0.80 mm in the best configuration. The proposed system achieves a sensitivity of 34.81% for 13 mm thick crystals and does not show a prediction degradation near the boundaries of the detector. We use the trained network to reconstruct an image of a grid of 21 point sources spread across the field-of-view and obtain a mean spatial resolution of 0.55 mm. We show that 25,000 Line of Responses (LORs) are needed to reconstruct a realistic OOC phantom with adequate image quality. Conclusions We demonstrate that it is possible to achieve a spatial resolution of almost 0.5 mm in a PET system made of multiple monolithic LYSO crystals by directly predicting the scintillation position from light patterns created with SiPMs. We observe that a thinner crystal performs better than a thicker one, that increasing the SiPM size from 3 mm to 6 mm only slightly decreases the prediction performance, and that certain surfaces encode significantly more information for the scintillation-point prediction than others.

Keywords

• CNN
• Deep learning
• GATE
• Monte-Carlo simulation
• Organs-on-chips
• PET