Preliminary Study Of The Semi-quantitative Accuracy And Image Quality Of ⁶⁸Ga-PSMA PET/CT With Different Beta Values And Algorithms

Objective:1)Based on block sequential regularized expectation maximization algorithm,to compare the effects of differentβ on the quantitative accuracy and image quality of ⁶⁸Ga-PSMA PET/CT after PVC reconstruction.2)This experiment compares the effects of different reconstruction algorithms on the semi-quantitative accuracy and image quality of ⁶⁸Ga-PSMA PET/CT.Methods:1)In the model experiment,image reconstruction is carried out in the range ofβ from 100 to 1000.RC,CR and BV are measured to evaluate semi-quantitative accuracy and image quality.In the clinical study,retrospectively collected image data of 21 patients who underwent ⁶⁸Ga-PSMA PET/CT examination at the First Hospital of Shanxi Medical University from March 2019 to February 2020,and a total of 29 abdominal imaging positive lymph nodes were collected,All included in the 100-1000β image reconstruction,according to the lymph node diameter measured under CT,divided into small lymph node group（diameter<10mm）and large lymph node group（diameter 10mm-30mm）.Two groups of SUVmax,SUVmean and SUVpeak were measured at the same time to perform PVC to evaluate the influence of differentβ on the two groups of SUV parameters.SNR and subjective scores were used to evaluate image quality.2)Except that the reconstruction method should be OSEM,TOF,TOF-PSF and BPL reconstruction,the rest of the uniform method one is the same.Use SPSS Statistics 25.0 for statistical analysis.Use the t test to analyze the quantitative data of the normal distribution.Cohen’s kappa coefficient is used to assess the consistency of physician scores.Results:1)The model experiment showed that with the increase ofβ ,CR,RC and BV all decreased.For clinical research,the image quality,image clarity,lesion significance, and total image scores given by nuclear medicine physicians all showed strong consistency.For the small lymph node group,theβ of 600 is the highest score;for the large lymph node group,theβ of 700 is the highest score.Theβ value from 100 to 600, the SNR of the two groups increased steadily.When theβ value is higher than 600,the SNR of the small lymph node group reaches a plateau,while the SNR of the large lymph node group continues to increase.The large lymph node group has a statistically significant difference in theβ between 600-1000（p<0.05）,while there was no significant statistical difference in the small lymph node group（p>0.05）.After PVC, the small lymph node group has more significant improvement in SUVs than the large lymph node group.Before PVC,there was significant negative correlation between all SUVs andβ values in the two groups.After conducting PVC,it was found that SUVmean and SUVmax of the two groups still had a significant negative correlation. However,the SUVpeak of the two groups showed no significant difference.2)According to OSEM,TOF,TOF-PSF and BPL algorithm for reconstruction,RC,CR increase.The RC and CR reconstructed using BPL reconstruction were significantly higher than other reconstruction methods（all P values<0.05）.The BPL reconstruction algorithm has the lowest BV.On the PET images reconstructed by different reconstruction algorithms in clinic,SUVmax and SUVmean are reconstructed according to OSEM,TOF,TOF-PSF to BPL gradually increased,and the SUVmax and SUVmean obtained under the BPL reconstruction algorithm were significantly higher than other reconstruction methods（P<0.001）.The SNR obtained under the regularization algorithm was significantly higher than other reconstruction methods（P <0.001）.Conclusion:1)Based on subjective scoring and semi-quantitative indicators,⁶⁸Ga-PSMA PET/CT is based on the block sequential regularized expectation maximization algorithm to select β of 600-700 for image reconstruction.The SUVpeak of small lesions after PVC is stable,and the clinical value should be explored further.2)Through the two parts of model and clinical research,⁶⁸Ga-PSMA PET/CT based on regularization algorithm has significantly improved semi-quantitative accuracy and image quality compared with other algorithms.