Virtual monoenergetic imaging trade-offs in postoperative spectral head CT: a two-center reader study

Postoperative non-contrast spectral head CT permits retrospective virtual monoenergetic imaging (VMI), but the energy level that best balances objective and reader-rated image quality is uncertain. We evaluated 93 patients who contributed 120 postoperative examinations using conventional images and VMIs at 60, 70, 80, 90, and 100 keV. Image noise, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and independent ratings by two neuroradiologists were analyzed with mixed-effects models reflecting the hierarchical data structure. CNR was highest at 60 keV (1.62 ± 0.39), whereas image noise was lowest at 100 keV (4.27 ± 0.49 HU). The 70-keV VMI had the highest reader scores for overall image quality, operative-bed parenchymal conspicuity, and diagnostic acceptability, whereas 100 keV had the highest artifact-suppression score. Quadratic-weighted kappa values ranged from 0.859 to 0.986. Thus, no single VMI energy optimized all image-quality properties: 60 keV maximized measured CNR, 70 keV produced the most favorable reader ratings across three routine-review domains, and 100 keV minimized noise and maximized artifact suppression. Diagnostic-performance validation is required before these findings inform a task-specific clinical workflow.