Spectral imaging for differentiating intrapancreatic accessory spleen from hypervascular pancreatic neoplasms using single-phase post-contrast CT.
To evaluate the diagnostic performance of spectral imaging for differentiating intrapancreatic accessory spleen (IPAS) from hypervascular pancreatic neoplasms using portal-venous phase (PVP) spectral CT.
This retrospective study included patients with hypervascular pancreatic lesions identified on PVP spectral CT. Lesion and splenic densities were measured on conventional images, virtual monochromatic images (VMI; 40-140 keV), virtual non-contrast (VNC) images, iodine maps, and effective atomic number (Z_eff) maps. Relative lesion-to-spleen ratios (RLSRs), with values closer to zero indicating greater similarity to the spleen, were compared between IPAS and neoplasms; diagnostic performance in the whole cohort and among patients with pancreatic tail lesions was assessed using the area under the receiver operating characteristic curve (AUC). For pancreatic tail lesions, three radiologists visually graded IPAS likelihood using conventional imaging alone and conventional imaging plus spectral imaging. Pooled sensitivity/specificity/accuracy and intraclass correlation coefficients (ICCs) were used to evaluate visual discriminative performance and inter-observer reproducibility, respectively.
Eighty-five patients (median age, 63.0 [IQR: 56.0-73.0] years; 45 men) with 34 IPAS and 51 neoplasms (34 IPAS and 22 neoplasms in the pancreatic tail) were included. RLSRs were lower in IPAS than in neoplasms across all image series (p < 0.001). In the whole-cohort/pancreatic-tail subgroup, AUCs of RLSR on conventional imaging, VNC, 40-keV VMI, iodine density, and Z_eff were 0.799/0.737, 0.835/0.899, 0.950/0.972, 0.960/0.980, and 0.969/0.992, respectively. In visual assessment, conventional plus spectral imaging improved pooled sensitivity/specificity/accuracy and ICC compared with conventional imaging alone (86.3%/90.9%/88.1% vs. 61.8%/59.1%/60.7%, 0.929 vs. 0.649, all p < 0.001).
Spectral imaging may facilitate differentiation of IPAS from hypervascular pancreatic neoplasms using only portal-venous-phase CT.
Question Can spectral imaging applied to hypervascular pancreatic lesions differentiate intrapancreatic accessory spleen (IPAS) from hypervascular neoplasms when only a single contrast-enhanced phase is available? Findings Use of spectral imaging yielded higher quantitative and qualitative performance for differentiating IPAS from neoplasms than conventional imaging, with better interobserver reproducibility in visual assessment. Clinical relevance Evaluating spectral properties may facilitate discrimination of IPAS from hypervascular pancreatic neoplasms identified on routine portal venous phase CT that would otherwise remain indeterminate on conventional CT and require additional diagnostic procedures.
This retrospective study included patients with hypervascular pancreatic lesions identified on PVP spectral CT. Lesion and splenic densities were measured on conventional images, virtual monochromatic images (VMI; 40-140 keV), virtual non-contrast (VNC) images, iodine maps, and effective atomic number (Z_eff) maps. Relative lesion-to-spleen ratios (RLSRs), with values closer to zero indicating greater similarity to the spleen, were compared between IPAS and neoplasms; diagnostic performance in the whole cohort and among patients with pancreatic tail lesions was assessed using the area under the receiver operating characteristic curve (AUC). For pancreatic tail lesions, three radiologists visually graded IPAS likelihood using conventional imaging alone and conventional imaging plus spectral imaging. Pooled sensitivity/specificity/accuracy and intraclass correlation coefficients (ICCs) were used to evaluate visual discriminative performance and inter-observer reproducibility, respectively.
Eighty-five patients (median age, 63.0 [IQR: 56.0-73.0] years; 45 men) with 34 IPAS and 51 neoplasms (34 IPAS and 22 neoplasms in the pancreatic tail) were included. RLSRs were lower in IPAS than in neoplasms across all image series (p < 0.001). In the whole-cohort/pancreatic-tail subgroup, AUCs of RLSR on conventional imaging, VNC, 40-keV VMI, iodine density, and Z_eff were 0.799/0.737, 0.835/0.899, 0.950/0.972, 0.960/0.980, and 0.969/0.992, respectively. In visual assessment, conventional plus spectral imaging improved pooled sensitivity/specificity/accuracy and ICC compared with conventional imaging alone (86.3%/90.9%/88.1% vs. 61.8%/59.1%/60.7%, 0.929 vs. 0.649, all p < 0.001).
Spectral imaging may facilitate differentiation of IPAS from hypervascular pancreatic neoplasms using only portal-venous-phase CT.
Question Can spectral imaging applied to hypervascular pancreatic lesions differentiate intrapancreatic accessory spleen (IPAS) from hypervascular neoplasms when only a single contrast-enhanced phase is available? Findings Use of spectral imaging yielded higher quantitative and qualitative performance for differentiating IPAS from neoplasms than conventional imaging, with better interobserver reproducibility in visual assessment. Clinical relevance Evaluating spectral properties may facilitate discrimination of IPAS from hypervascular pancreatic neoplasms identified on routine portal venous phase CT that would otherwise remain indeterminate on conventional CT and require additional diagnostic procedures.
Authors
Nagayama Nagayama, Inoue Inoue, Ishiuchi Ishiuchi, Yamamura Yamamura, Igarashi Igarashi, Kidoh Kidoh, Oda Oda, Nakaura Nakaura, Hirai Hirai
View on Pubmed