Physical characteristics of deep learning-based image processing software in computed tomography: a phantom study

Purpose This study aimed to assess the image characteristics of deep-learning-based image processing software (DLIP; FCT PixelShine, FUJIFILM, Tokyo, Japan) and compare it with filtered back projection (FBP), model-based iterative reconstruction (MBIR), and deep-learning-based reconstruction (DLR). Methods This phantom study assessed the object-specific spatial resolution (task-based transfer function [TTF]), noise characteristics (noise power spectrum [NPS]), and low-contrast detectability (low-contrast object-specific contrast-to-noise ratio [CNR LO ]) at three different output doses (standard: 10 mGy; low: 3.9 mGy; ultralow: 2.0 mGy). The processing strength of DLIP FBP with A1, A4, and A9 was compared with those of FBP, MBIR, and DLR. Result The standard dose with high-contrast TTFs of DLIP FBP exceeded that of FBP. Low-contrast TTFs were comparable to or lower than that of FBP. The NPS peak frequency ( f P ) of DLIP FBP shifts to low spatial frequencies of up to 8.6% at ultralow doses compared to the standard FBP dose. MBIR shifted the most f P compared to FBP—a marked shift of up to 49%. DLIP FBP showed a CNR LO equal to or greater than that of DLR in standard or low doses. In contrast, the CNR LO of the DLIP FBP was equal to or lower than that of the DLR in ultralow doses. Conclusion DLIP FBP reduced image noise while maintaining a resolution similar to commercially available MBIR and DLR. The slight spatial frequency shift of f P in DLIP FBP contributed to the noise texture degradation suppression. The NPS suppression in the low spatial frequency range effectively improved the low-contrast detectability.