Inter-software agreement in abdominal computed diffusion-weighted imaging: apparent diffusion coefficient and image quality assessment.

PURPOSE

To investigate the agreement in apparent diffusion coefficient (ADC) values and the image quality of abdominal computed diffusion-weighted imaging (cDWI) generated by two software programs.

METHODS

This retrospective study included 154 patients (64 males, 90 females; mean age: 70.1 ± 10.7 years) who underwent abdominal diffusion-weighted imaging (DWI) between November 2023 and October 2024. DWI was acquired with a b = 0 and 1000 s/mm using a 3.0T scanner. cDWIs with b 1500 and 2000 were generated by a scanner console software and a commercially available software program. The ADC and signal-to-noise ratio (SNR) were measured in normal tissues. The contrast-to-noise ratio (CNR) of normal tissues relative to the paraspinal muscle was also measured. Agreement and reliability of ADC value were assessed using Bland-Altman analysis and intraclass correlation coefficients (ICC), respectively. SNR and CNR were compared using paired t-tests.

RESULTS

ADC bias and limits of agreement ranged from - 0.2 to 0.7% and 0.3-2.1%, respectively. ICCs for ADC values ranged within 0.991 and 1.000 (p < 0.05) for all tissues. Significant SNR differences were observed in b1500 (liver, spleen, kidney) and b2000 (liver, bile). Significant CNR differences were noted in b1500 (liver, pancreas, kidney) and b2000 (excluding spleen).

CONCLUSION

There is strong agreement and high reliability in ADC values between the two software programs. Although significant differences in SNR and CNR were observed, their clinical impact appears minimal. These findings indicate that the ADC values and image quality of cDWI are consistent between the two software programs.