Modified NEMA NU-2 performance evaluation methods for a brain-dedicated PET system with a hemispherical detector arrangement.

Brain PET imaging has important roles in neurology, neuro-oncology and molecular imaging research. We have developed a helmet-type PET prototype and have shown that the proposed hemispherical geometry had high potential for realizing high-sensitivity and low-cost brain imaging. However, there is no standard performance evaluation method for helmet-type PET, which would be a bottleneck to its commercialization. Therefore, we investigated appropriate performance evaluation methods for a helmet-type PET based on the NEMA NU 2-2018 standards. For those measurement methods that are not applicable to the helmet-type PET, we changed them while keeping the basic concept of the original NEMA standards. We measured spatial resolution, sensitivity, scatter fraction, count rate characteristics, accuracy of corrections for count losses and randoms, and image quality. We partially changed the measurement methods by making brain-size phantoms and by optimizing the length or the position of radioactive sources. The spatial resolution was 2.8 mm at 1-cm offset position by the filtered back-projection method. Sensitivities measured by the NEMA original setup and the proposed setup were 13.4 and 57.1 kcps/MBq. The respective values measured with our developed brain-size scatter phantom and with the conventional whole-body-size scatter phantom were: scatter fractions of 35% and 35%; peak NECRs of 25.1 kcps at 3.2 kBq/ml and 19.8 kcps at 2.6 kBq/ml ; and maximum absolute biases of 5.5% and 16.0%. The image quality was evaluated with the developed brain-size phantom, and good image quality was obtained. The helmet-type PET prototype showed high-sensitivity even with the small number of 54 detectors. The spatial resolution was better than 4.0 mm over the field-of-view. In conclusion, we proposed the performance evaluation methods for a brain-dedicated PET system with a hemispherical geometry. The proposed method could facilitate evaluation of performance characteristics of brain-dedicated PET scanners and optimization of its scanning and reconstruction parameters.