OPTIMIZATION OF PHOTON-COUNTING CT FOR BONE IMAGING USING ULTRA-HIGH-RESOLUTION MODE

While ultra-high-resolution (UHR) photon-counting detector CT (PCD-CT) shows promising agreement with high resolution peripheral quantitative CT for bone microarchitecture evaluation, clinical translation requires balancing noise, resolution, and dose. Conventional optimization relies on objective metrics and may not reflect task-based perception. A hybrid approach integrating objective and subjective assessment may better guide the development of clinically robust PCD-CT UHR bone imaging protocols.

To apply a hybrid objective-subjective approach to optimize PCD-CT UHR bone imaging protocol at clinically acceptable dose levels.

Four cadaveric specimens (one lumbar vertebra and three proximal femora) along with the Mindways phantom (Model 3), were imaged with PCD-CT (NAEOTOM Alpha.Peak VB20, Siemens Healthineers) at 120kVp. Six total acquisitions were performed at three dose levels, IQ-75 (3.0-4.3mGy), IQ-150 (6.0-8.6mGy), and IQ-450 (17.9-25.7mGy) across two pitch settings (0.35 and 0.85). For each acquisition, images were reconstructed using three kernels (Br76, Br89, Br98) and two levels of Quantum Iterative Reconstruction (QIR2 and QIR4), yielding 35 unique protocols. Protocol performance was evaluated using both objective and subjective assessment. Objective metrics included SNR, CNR, and task-based transfer function (TTF) f50 for spatial resolution. Task-based protocol performance was independently assessed by four radiologists using a 5-point Likert scale for cortical sharpness, trabecular visibility, and noise. Statistical analysis for SNR, CNR, TTF-f50 and Likert ratings were performed, and heatmaps were generated to evaluate protocol performance. Interrater reliability was assessed using a two-way random-effects, absolute-agreement intraclass correlation coefficient (ICC).

SNR and CNR were highest with Br76 kernel, QIR4, and highest exposure (IQ-450), while noise was lowest for Br76, increased with sharper kernels (Br89, Br98) and lower IQ, and was markedly reduced with QIR4; pitch had minimal impact (figures 1, 2). Variance analysis showed that QIR had the greatest impact (η² = 0.54, 0.52), followed by kernel (η² = 0.29, 0.32) and IQ (η² = 0.09, 0.08), while pitch had no significant effect. For Br76, TTF-f50 was consistently higher with QIR2 than QIR4 (1.62 vs 1.36 at IQ-75, pitch 0.85), with minimal dependence on IQ and pitch. Consistently, Br76 achieved the highest scores (mean Likert 4.2), outperforming Br89 (3.88) and Br98 (3.02), with scores improving with increasing IQ and at QIR2, and minimal impact from pitch. Interrater reliability was excellent (ICC = 0.92).

Br76 provided the best objective-subjective performance for trabecular bone assessment, with QIR4 minimizing noise and maximizing SNR/CNR, while QIR2 better preserved spatial resolution; pitch had no significant effect. Protocol optimization of PCD-CT for UHR imaging of trabecular bone at a clinically acceptable radiation dose is essential to unlock its full potential in clinical research.