To investigate the feasibility of virtual monoenergetic images (VMIs) derived from a dual-layer detector spectral CT (SDCT) in patients with breast cancer by assessing tumor conspicuity in comparison with conventional images (CI) and to correlate tumor conspicuity on VMIS with prognostic biomarkers.
64 patients with pathologically proven benign or breast cancers (14 benign lesions and 65 breast cancers) underwent arterial and 90s delayed phase scan on a SDCT for tumor staging between June 2016 and May 2018. CI were reconstructed at 120kVp, and VMI at 40 keV (VMI40). A retrospective spectral data analysis was performed to assess the regions of interest by using CI and VMI40 on arterial and delayed images (CIART, VMI40ART, CIDE, and VMI40DE). Two radiologists independently reviewed the 4 image sets for the conspicuity score of breast lesions. For quantitative analysis, measurement of Hounsfield units (HU) of breast lesions on VMI40ART and VMI40DE was performed. Receiver operating characteristic (ROC) analysis was performed. Estrogen receptor (ER), human epidermal growth factor receptor 2 (HER2), and Ki67 level were evaluated using histopathology. Statistically, conspicuity score and mean HU of malignant lesions were correlated with histological characteristics.
The mean conspicuity score of malignant lesions was significantly higher compared with that of benign lesions in all image sets (P < 0.05). The area under the ROC curve of conspicuity score for malignant lesions were greatest on VMI40DE. VMI40DE yielded significantly higher mean HU of malignant lesions compared to VMI40ART (P < 0.001). Malignant lesions on VMI40ART and VMI40DE revealed significantly higher mean HU compared to those of benign lesions (P < 0.001). The conspicuity score and the mean HU on VMI40ART were significantly higher in cancers with diameter greater than 2 cm, ER negativity, PR negativity, and Ki67 positivity (P < 0.05).
VMI40DE is feasible to diagnose breast cancers with higher conspicuity score and better contrast enhancement compared with VMI40ART. Moreover, VMI40ART may serve as additional predictors of a poor breast cancer prognosis.
Clinical trial identification
Legal entity responsible for the study
Has not received any funding.
The author has declared no conflicts of interest.