Initial exploration of coronary stent image subtraction using dual-layer spectral CT

This study aimed to investigate the feasibility of coronary stent image subtraction using spectral tools derived from dual-layer spectral computed tomography (CT). Forty-three patients (65 stents) who underwent coronary CT angiography using dual-layer spectral CT were included. Conventional, 50-keV (kilo electron-volt), 100-keV, and virtual non-contrast (VNC) images were reconstructed from the same cardiac phase. Stents were subtracted on VNC images from conventional (conv${}_{sub}$), 100-keV (100-keV${}_{sub}$), and 50-keV (50-keV${}_{sub}$) images. The in-stent lumen diameters were measured on subtraction, conventional, and 100-keV images. Subjective evaluation of reader confidence and subtractive quality was evaluated. Friedman tests were performed to compare in-stent lumen diameters and subjective evaluation among different images. Correlation between stent diameter and subjective evaluation was expressed as Spearman's rank correlation coefficient (r${}_s$). The diagnostic accuracy was assessed according to invasive coronary angiography (ICA) performed in 11 patients (20 stents). In-stent lumen diameters were significantly larger on subtraction images than those on conventional and 100-keV images (p < 0.05). Higher reader confidence was found on 100-keV, conv${}_{sub}$, 100-keV${}_{sub}$, and 50-keV${}_{sub}$ images compared with conventional images (p < 0.05). Subtractive quality of 100-keV${}_{sub}$ images was better than that of conv${}_{sub}$ images (p = 0.037). A moderate-to-strong correlation between stent diameter and subjective evaluation was found (r${}_s$ = 0.527~0.790, p < 0.05). Higher specificity, positive predictive value, and negative predictive value of subtraction images were shown by ICA results. Subtraction images derived from dual-layer spectral CT enhanced in-stent lumen visibility and could potentially improve diagnostic performance for evaluating coronary stents.