Objective To investigate the value of deep learning reconstruction in improving image quality of sacroiliac joint CT in overweight patients.Methods A retrospective analysis was performed on 118 overweight(BMI ≥ 24 kg/m2)patients who underwent sacroiliac joint CT in our hospital between March 2017 and May 2019.The patients were randomly assigned to standard-dose CT(SDCT),low-dose CT(LDCT),and ultra-low-dose CT(ULDCT)groups.Images of SDCT were reconstructed with hybrid iterative reconstruction(HIR)algorithm whereas images of LDCT and ULDCT were reconstructed with deep learning reconstruction(DLR)algorithm.Radiation exposure,image noise,signal-to-noise ratio(SNR)and contrast-to-noise ratio(CNR)of the first sacral vertebra were compared.Subjective image quality and visualization of sacroiliitis radiographic features were evaluated based on 5-point scales.The radiation dosage,objective and subjective image quality of the 3 groups were compared using one-way or Kruskal-Wallis ANOVA test.Results The effective radiation doses of LDCT[(1.01±0.07)mSv]and ULDCT[(0.43±0.02)mSv]were significantly(P＜0.001)reduced by 32.2％and 71.1％than that of SDCT[(1.49±0.10)mSv],respectively.Image noise,SNR and CNR showed significant(P＜0.001)differences among the 3 groups and in the pairwise comparisons.LDCT demonstrated significantly(all P＜0.05)lower noise(25.05±2.75),higher SNR(10.38±0.56)and CNR(7.92±0.50)compared with ULDCT(31.26±3.51,8.27±0.60,6.71± 0.49)and SDCT(51.25±1.59,4.70±0.23,3.55±0.20).The overall image quality of LDCT was significantly higher than SDCT(P=0.001)and ULDCT(P=0.018)without significant difference between SDCT and ULDCT(P=0.364).DLR-LDCT was significantly better than HIR-SDCT for visualizing bone erosion,joint space narrowing or widening and subarticular cysts(all P＜0.05).Conclusion DLR improves image quality and effectively reduces radiation exposure.DLR-LDCT is better than HIR-SDCT for evaluating sacroiliitis in overweight patients.
维普
万方数据