Deep learning reconstruction-applied PROPELLER diffusion-weighted imaging in uterine malignancies: a comparison study with MUSE.

PURPOSE: To compare the image quality of diffusion-weighted imaging (DWI) between deep learning reconstruction (DLR)-applied Periodically Rotated Overlapping ParallEL Lines with Enhanced Reconstruction (PROPELLER) and MUltiplexed Sensitivity Encoding (MUSE) sequences in MRI of uterine malignancies. MATERIALS AND METHODS: This retrospective study included 66 MRI examinations of 48 patients. The cohort consisted of cases with uterine malignancies (cervical cancer (n = 50), endometrial cancer (n = 15) and endometrial stromal sarcoma (n = 1)) at different clinical settings (initial staging, post-treatment, or recurrence). DWI was performed using both DLR-applied PROPELLER and MUSE sequences using b-values of 0 and 800 s/mm2. Two independent reviewers evaluated overall image quality, artifacts, sharpness, and lesion conspicuity using a 5-point Likert scale. Apparent diffusion coefficient (ADC), apparent signal-to-noise ratio (aSNR), and apparent contrast-to-noise ratio (aCNR) were quantitatively measured in both sequences. RESULTS: In the qualitative assessment, DLR-applied PROPELLER-DWI showed significantly fewer artifacts and better lesion conspicuity than MUSE-DWI. No significant difference was found in overall image quality or sharpness. Inter-reader agreement was slight to moderate, except for strong agreement on DLR-applied PROPELLER-DWI lesion conspicuity (κ = 0.696, good). Quantitatively, the mean ADC values were significantly higher with DLR-applied PROPELLER-DWI compared to MUSE-DWI for both readers (p < 0.001). Furthermore, both aSNR and aCNR were significantly higher with DLR-applied PROPELLER-DWI compared to MUSE-DWI. CONCLUSIONS: DLR-applied PROPELLER-DWI demonstrated fewer artifacts and better lesion conspicuity than MUSE-DWI, with higher measured aSNR and aCNR. However, DLR-applied PROPELLER-DWI yielded significantly higher mean ADC values.