Performance of detector modules designed for the NeuroSphere brain PET system

The Human Dynamic NeuroChemical Connectome (HDNCC) scanner integrates a high spatio-temporal resolution dedicated brain PET insert (NeuroSphere PET) with a commercial 7-T MRI to enable studies of neurochemical dynamics simultaneously with functional MRI on comparable timescales.

The NeuroSphere will achieve unprecedented sensitivity with a spherical geometry and novel detectors, which must provide time-of-flight (TOF) and depth-of-interaction (DOI) encoding without compromising crystal separation or energy resolution. We aim to develop a novel PET module for high-resolution PET/MR applications with TOF and DOI capability.

The NeuroSphere PET detector is a 10 × 10 array of 26.0-mm LSO crystals with 1.6-mm pitch coupled to a 4 × 4 SiPM array with 4-mm pitch, and a top light guide that provides DOI encoding via light sharing. Time-to-digital converter (TDC) performance and nonlinearity correction methods were evaluated with a pulsed laser. Crystal separation, energy resolution, and DOI resolution were evaluated, including a DOI-based energy correction. The ASIC readout parameters were configured and timing corrections were implemented to optimize coincidence timing resolution (CTR), including a DOI-based timing correction.

The TDC achieved 55-ps CTR between laser pulses after implementing nonlinearity correction, an improvement to the factory correction. The PET detector had good crystal identification with a peak-to-valley ratio of 11.43 $\pm$ 5.87. DOI resolution was 7.45 $\pm$ 2.02 mm full-width-half-maximum (FWHM). A batch of 14 detector blocks had an energy resolution of 11.0% $\pm$ 1.8 % FWHM, improved from 12.6% $\pm$ 1.7 % FWHM by a DOI-based correction. The best block-to-block CTR was found to be 510 ps FWHM, with $>$ 100 ps improvement after a DOI-based correction, and a batch of 7 block pairs had a CTR of 526 $\pm$ 15 ps FWHM.

This work has shown the capability of the NeuroSphere PET module to provide energy resolution, DOI encoding, and TOF PET competitive with other dedicated brain PET/MR systems. The CTR of $\sim$ 500 ps provides SNR gain for neuroimaging, and the energy resolution and CTR were improved by a DOI-based correction using an easily scalable calibration procedure.