光子計数検出器CTを用いた側頭骨イメージングにおける再構成カーネルの最適化：物理的および視覚的評価の組み合わせ

Photon-counting detector CT (PCD-CT) offers superior spatial resolution and noise characteristics compared to conventional CT. However, optimal reconstruction parameters for temporal bone imaging, especially kernel selection, remain unclear. This study aimed to identify the optimal reconstruction kernel using both objective physical image quality metrics and subjective expert assessments.

In phantom experiments, the system performance function (SPF) based on the task-based transfer function (TTF) and noise power spectrum (NPS) was calculated across 11 reconstruction kernels (Hr60-Hr98). Based on the results of the physical evaluation and clinical considerations from clinical practice, a subset of kernels was selected for visual assessment. For clinical images, two diagnostic radiologists evaluated three fine anatomical structures (i.e., stapes footplate, incudomalleolar joint, and cochlea) and overall image quality using both a ranking method and a 5-point Likert scale.

TTF analysis indicated that Hr96 had the highest spatial resolution, while Hr60 showed the lowest noise in the NPS. SPF analysis identified Hr72 as providing the optimal balance between resolution and noise. Visual assessment using four reconstruction kernels (Hr60, Hr72, Hr76, and Hr84) showed that Hr76 consistently received the highest preference for overall image quality and visualization of fine structures. Statistically significant differences were observed among the kernels, with Hr60 consistently rated the lowest (p < 0.05).

The kernel Hr76 was found suitable for middle and inner ear diagnoses using PCD-CT, providing a good balance between spatial resolution and image noise. This finding provides a foundation for standardized reconstruction protocols in high-resolution temporal bone imaging.

These findings support the use of Hr76 as a standard kernel for high-resolution temporal bone imaging and may contribute to protocol optimization in clinical PCD-CT practice.