This study combined physical phantom testing of 11 UHR reconstruction kernels (Hr60–Hr98) with clinical visual assessment in 12 patients / 24 temporal bones. Physically, increasing kernel sharpness improved spatial resolution but increased high-frequency noise: Hr96 achieved the highest measured resolution, whereas Hr60 produced the lowest noise.
The objective system-performance analysis favored Hr72 as the best resolution–noise compromise. However, expert visual assessment of Hr60, Hr72, Hr76, and Hr84 consistently favored Hr76, particularly for visualization of the cochlea and fine middle-ear anatomy; Hr60 performed worst despite its low noise.
The practical message is that maximum sharpness is not equivalent to maximum diagnostic image quality. For UHR temporal-bone PCCT at 0.2-mm reconstruction, Hr76 appears to provide the most useful clinical balance, while Hr72 remains a reasonable slightly smoother alternative.
Access publication
Figure 5. Typical CT images used to evaluate the incudomalleolar joint. (a), (b), (c), and (d) correspond to the images reconstructed using Hr60, Hr72, Hr76, and Hr84, respectively. Sharper kernels enhanced the visualization of fine structures.
Protocol
Detail | Value |
|---|
Scanner | NAEOTOM Alpha |
Scan area | Temporal bone |
Acquisition mode | Quantum HD (UHR) |
Tube voltage [kV] | 120 |
Effective mAs | Phantom: 305; clinical cohort: 266.1 ± 18.1 |
IQ level | Clinical cohort: 280 |
Dose modulation | Automatic exposure control |
CTDIvol [mGy] | Phantom: ~52; clinical cohort: 45.63 ± 3.11 |
Rotation time [s] | Phantom: 0.50 |
Pitch | Phantom: 0.55 |
Detector slice collimation | 120 × 0.2 mm |
FOV [mm] | Phantom: 100 × 100 |
Matrix | 512 × 512 |
Slice thickness | 0.2 mm |
Reconstruction increment | 0.2 mm |
Iterative reconstruction | QIR 1 (to preserve high-frequency information) |
Spectral reconstruction | T3D polychromatic images |
Kernels physically tested | Hr60, Hr64, Hr68, Hr72, Hr76, Hr80, Hr84, Hr89, Hr92, Hr96, Hr98 |
Kernels visually tested | Hr60, Hr72, Hr76, Hr84 |
Clinical evaluation structures | Stapes footplate, cochlea, incudomalleolar joint, overall image quality |
Clinical image review | 24 bilateral temporal-bone image sets; 2 radiologists |
Key Results
Parameter | Main finding |
|---|
Highest spatial resolution | Hr96, 10% TTF 2.712 cycles/mm |
Lowest image noise | Hr60 |
Best physical resolution–noise balance | Hr72, highest SPF |
Best visually rated kernel | Hr76 overall |
Cochlea ranking | Hr76 ranked highest; significantly better than Hr84 |
Inter-reader agreement | ρ 0.634–0.733 for ranking and 0.834–0.939 for Likert ratings |
Source:
Nishii S, Asahara T, Morimitsu Y, Kajisaki S, Akagi N, Honda M, Hayashi H, Sugaya A, Munetomo K, Higaki F, Hiraki T, Iguchi T. Optimization of the reconstruction kernel for temporal bone imaging using photon-counting detector CT: A combined physical and visual evaluation. Radiography. 2026;32(2):103274. doi:10.1016/j.radi.2025.103274.
