Anemia is a hematologic condition characterized by a reduction in the number of circulating red blood cells (RBC) or their Hb content, diagnosed through estimation of Hb concentration usually in the laboratory [33, 34]. Anemia detection in CT datasets has the potential to identify clinically silent anemia, prompting further laboratory and clinical investigations.

We assumed a correlation with the atomic number Zeff. If Zeff is measured in a material consisting of only one element, it reflects the atomic number. For mixed materials, Zeff is determined by the respective elements and their number of electrons, resulting in an averaged value that cannot be assigned to a specific atomic number. In DECT, Zeff is determined using a coefficient that is calculated from previous calibration data, including ρ measurements, which, in turn, was considered within the utilized application profiles [14]. Zeff reflects the blood composition on an atomic level. Hb, an iron-containing oxygen-transport protein composed of four heme molecules and one globin protein, is present in erythrocytes up to about 96% of a red blood cell’s dry weight [35,36,37]. Even if quantitative changes in the number of Hb proteins might alter the mean Zeff values, the results of our study show no positive or negative correlation measurable using DECT (refer to Table 2).

It appears that anemia has no significant or discernible effect on the overall elemental blood composition on an atomic level measurable in CT, despite changes in RBC concentration. Overall, Zeff seems to be a good parameter for determining single elements or differentiating selective materials, but has limitations in complex liquid materials like blood [6, 12, 13, 38].

By using two energy spectra, the electron density relative to water (ρ, rho), comparable to the attenuation measurements in SECT, can be determined in DECT as well [15]. To our knowledge, we used ρ for the first time in the literature for a linear regression analysis between Hb and CT measurements. Therefore, we demonstrated a positive correlation between Hb and ρ (refer to Table 2). As the solid blood components mainly consist of RBC, and the hematocrit indicates the percentage of cellular blood components, an additional positive correlation between ρ and hematocrit was observed, as expected.

While ρ showed a positive correlation with the Hb/Hct value, within the scope of our analyses, the strongest correlation was observed for attenuation (refer to Table 2). This was evident in both tbDECT and dsDECT with overall better results for tb acquisition (refer to Fig. 4). Better correlation for attenuation may be because ρ measurements are based on calibration scans and calculations from calibration coefficients that depend on low- and high-energy DECT images [14].

Consequently, numerous studies have investigated the relationship between CT measurements and Hb/Hct levels before, establishing a positive correlation between the blood and attenuation values [16, 19, 21,22,23,24]. However, we observed some important limitations for some of them. First, we noticed variations in the scan protocols that may significantly affect attenuation measurements in SECT [4, 5]. Title et al, for example, used two different CT scanners with different tube voltages (120 kV and 140 kV) for their study population, and Chaudhry et al reported a peak tube voltage of a maximum of 120 kV [19, 22]. Jung et al and Zhou et al used care dose implementations like automatic tube current, with tube current values ranging from 30 milliampere seconds (mAs) up to 500 mAs [20, 21]. Also, physical patient variation like different body weights, influencing care dose implementations, was unavoidable with the in vivo study designs [26]. Additionally, the time delay between the CT examination and the laboratory determination of Hb/Hct is an important limitation, since the blood values are not stable parameters in living patients. Depending on the study, time latency was reported to range from hours, as in the study from Title et al, up to 7 days, as Zopfs et al used as inclusion criteria [21, 23, 24, 26].

Our approach with an in vitro study eliminates these influencing factors at once. On the one hand, DECT measurements are known to be more valid an show less susceptibility to changes in the scan protocol [1, 15]. On the other hand, we overcome the confounding factor of changing scan parameters through a standardized scan protocol (refer to Table 1). The standardized scan setting allowed us to exclude patient-related confounding factors. The fact that the Hb level is relatively stable in properly stored blood samples leads to a theoretical time delay between laboratory evaluation and a CT scan of zero [39, 40].

Nevertheless, the in vitro setting resulted in the limitation that, despite thorough agitation of the blood samples, a minor influence of the sedimentation rate cannot be ruled out. The study was conducted using one of the latest single-source CT devices available on the market, despite dual-source DECT devices representing the current state-of-the-art. It should be noted that this decision may slightly limit the comparison with clinical practice. A comparison between tb and ds acquisition techniques as possible; however, despite the standardized scan protocol, it should be noted that the tube voltage differed between the two scan modes, and therefore, the comparison must be critically evaluated. The superior results obtained with tbDECT, as mentioned before, remain inconclusively explained, but, besides the differed utilized tube voltage, a plausible reason may lie in the scanning procedure. While tbDECT requires a single scan to generate low- and high-energy images, dsDECT necessitates two scans for the same purpose. Although only a few seconds elapse between these scans, this latency may noticeably affect erythrocyte sedimentation within the measurement results. Another relevant limitation of our in vitro study is the grid arrangement we utilized, which proved to be the most practical for tilting the numerous blood samples and preventing blood sedimentation. The resultant lack of dense material surrounding the blood samples and the absence of beam hardening effects may restrict the comparability with in vivo studies.

Some studies have determined cutoff values for anemia detection using attenuation values as a predictor [21, 26]. Our aim with this in vitro study was to complement existing in vivo studies and provide a scientifically grounded foundation for future research in this area. Nevertheless, our statistical analysis indicates good differentiation between blood samples with and without anemia within the use of CT measurements (refer to Figs. 5 and 6). We identified several gender-specific ρ and attenuation cutoff values for ds and tb DECT. Given the better correlation between attenuation and Hb values, observed in our study while using DECT, which has also been shown in further studies with virtually non-contrast CT images or photon-counting CTs, attenuation seems to be the preferred parameter for computed tomographic anemia detection [25, 41]. But this, of course, does not replace the comprehensive laboratory analysis and anemia diagnostics and only can be used as a screening parameter.

In summary, our study investigated the correlation between DECT measurements and blood parameters, emphasizing novel aspects of ρ and Zeff values. The use of DECT offers less variation based on scan protocols and scanning blood samples addresses previous time latency issues. If quantitative changes in the Hb count might alter the mean Zeff values, the results of our study show that there is no measurable correlation using DECT. However, we conducted a linear regression analysis between Hb/Hct and ρ for the first time, demonstrating a positive correlation and indicating the potential of this parameter in further in vivo studies. Nevertheless, attenuation emerged as the most strongly correlated parameter with identifiable cutoff values, especially in tb DECT, highlighting its preference for CT-based anemia detection.