Our results showed that AEF-ID had excellent diagnostic value in differentiating between high- and low-grade BCa, demonstrating that the use of the spectral parameter AEF-ID to evaluate BCa pathology in diagnosis was effective and instructive. In addition, negative results indicated that other parameters might not be correlated with the pathology of BCa. To the best of our knowledge, this is the first study to evaluate BCa pathology using quantitative parameters obtained with DLCT.

In our multivariable analysis, the OR for AEF-ID was 1.089, with a 95% CI of 1.029–1.153. Although the OR value represented only a slight increase, the narrow CI and the p-value of 0.03 indicated that this increase was statistically significant. This suggested that AEF-ID was a meaningful predictor of BCa grade, despite the modest effect size. The narrow CI, which did not include the value 1.000, further underscored the robustness of this association in statistical terms. Given the small sample size in our study, we acknowledge that the observed effect might have been influenced by this limitation. Therefore, the slight increase in the OR value might underestimate AEF-ID’s true predictive power. To address this, we plan to conduct further research with a larger study population to more rigorously assess AEF-ID as a prognostic indicator of BCa grade.

The identified threshold of AEF-ID stood at 67.70%, accompanied by Sen of 95.5%, Spc of 81.0%, and AUC of 0.924. A higher AEF-ID value generally signifies a higher grade of BCa; Fig. 3a illustrates the distribution of the AEF value in both groups. Following are some possible explanations of why AEF performed better than other spectral CT parameters. First, from a hemodynamic perspective, infiltration of cancer cells in BCa induces angiogenesis [26, 27]. This leads to an augmented distribution of contrast agents in the intravascular/extravascular compartment, resulting in an elevated AEF value in the corresponding circulatory phase [28]. Moreover, the use of spectral detector CT in this study enabled concurrent capture of low- and high-energy data at identical spatial and temporal coordinates, resulting in impeccably aligned data and notably diminished measurement error without requiring any pre-determined selection for acquisition mode [29].

We also found good collinearity between AEF-ID and AEF-HU. DLCT, using dual- or multi-energy techniques, can more accurately separate and quantify iodine content in tissues, providing more precise calculations of AEF. However, traditional CT can also be used to obtain AEF values, suggesting that it can be used for analysis in the absence of advanced technologies such as spectral CT. Similar studies, such as that by Huber et al, have found that using AEF based on conventional CT can help screen for hepatocellular carcinoma; Huber et al suggested a cutoff AEF value of 50% [29].

Previous research has suggested that other spectral parameters can serve as valuable tools for assessing different types of cancer. For instance, Fujita et al discovered that ECV fraction could be used to help predict the efficacy of preoperative neoadjuvant chemotherapy in pancreatic ductal adenocarcinoma, which can be attributed to its association with the histological degree of fibrosis and quantity of desmoplastic stroma [30]. We hypothesized that ECV fraction could help differentiate between high- and low-grade BCa. In our univariate analysis, ECV fraction showed significance, while in multivariable analysis, it was not a significant parameter. This indicated that ECV fraction might lack the requisite statistical strength to be considered a robust differentiating factor for BCa grade in a more comprehensive prognostic model.

VNC images can provide baseline CT values just like true non-contrast images, which can differentiate tumors from surrounding normal tissues. The distinct appearances of tumors in pre- and post-contrast images can be more clearly revealed by VNC imaging, aiding in tumor identification and staging. Zhang et al showed that VNC images can maintain good image quality, permitting a decrease in radiation dose in the diagnosis of renal cell carcinoma [31]. Wang et al found that Zeff has high Sen and Spc for predicting pathological subtype and risk stratification of ground-glass nodules (GGNs), perhaps because the pathological subtypes of different GGNs are composed of different substances and can be reflected by Zeff [32]. Tanaka et al found that low-ID tumor area ratio was a useful prognostic index of non–small-cell lung cancer after stereotactic body radiotherapy [33], possibly due to tumor tissues typically having abnormal hemodynamic perfusion characteristics. Malignant tumors are often neovascularized, leading to increased blood flow. ID can help identify these differences in perfusion and assess the blood supply to the tumor.

Additionally, λHU reflects how the attenuation of various substances varies with the energy distribution of X-ray photons. This variation is linked to the composition and density of the substances, as well as to interactions between the photon energies and the substances themselves [25]. Unfortunately, our study did not identify the λHU as a strong predictor of BCa pathology. Further analysis of BCa’s cellular composition and pathological aspects might be necessary to clarify the underlying reasons. Alternatively, the negative results we observed could be ascribed to the limited sample size.

Our study had some limitations. First, due to circumstances beyond our control, our access to the Philips IQon CT scanner was limited to the period from October 2017 to October 2020 for routine CT scanning in our department; hence, this was a single-center exploratory study with a small sample size. Second, the spectral CT features we examined were not comprehensive, nor were they combined with conventional CT features. Certain crucial features might have been omitted, thereby necessitating future investigations to encompass a broader range of CT features for analysis. Third, not all pathological sections might have been matched with imaging ROIs, potentially introducing bias into the analysis.

In summary, the spectral parameter AEF-ID proved to be a significant prognostic factor in BCa grading, despite its modest effect size. We plan to validate its predictive value in a larger study population in subsequent studies. This approach could help clinicians devise initial treatment strategies, potentially improving patient outcomes.