Efficacy of dual-layer spectral detector computed tomography for detecting early ischemic changes in patients with acute ischemic stroke: A pilot study

INTRODUCTION

Acute ischemic stroke (AIS) is a critical neurological condition with significant health and economic impacts due to its high morbidity and mortality rates.^[1] Early detection and intervention are crucial because they can greatly improve patient outcomes by enabling timely thrombolytic or endovascular therapies.^[2] At present, non-contrast brain computed tomography (NCCT) is the primary imaging modality for emergency evaluation of acute stroke due to its rapid availability and ease of use.^[3] However, NCCT often lacks sensitivity in identifying subtle early ischemic changes (EICs), such as hypo-attenuation or loss of the gray-white matter interface, both of which are vital for prompt AIS diagnosis. In normal brain tissue, the density difference between the gray and white matter regions is typically 5–10 Hounsfield units (HU). In the early stages of cerebral ischemia, this difference decreases to <5 HU, resulting in a very low contrast between the normal and affected tissues, which makes lesion detection challenging.^[4,5] Advancements in CT technology, notably dual-layer spectral detector CT (DLCT), provide improved tissue characterization capabilities through spectral imaging, potentially addressing previous limitations in lesion detection. DLCT enables acquisition of both conventional and spectral images within a single scan by capturing spectral information from different photon energy levels. The derived parameters, such as the electron density (ED) and effective atomic number (effective Z), provide additional quantitative insights into the tissue composition and cellular changes.^[6,7] These advanced CT imaging techniques may aid in detecting subtle EICs associated with cellular edema, which are challenging to identify by conventional NCCT; however, their effectiveness in AIS has not yet been investigated.

This study aimed to assess the efficacy of DLCT, with a particular focus on ED and effective Z, for detecting EICs in patients with AIS. In addition, it explores the potential of DLCT as a more accessible alternative to magnetic resonance imaging (MRI) in emergency clinical settings.

MATERIAL AND METHODS

RESULTS

Quantitative image analysis

Statistically significant differences were identified in CT (28.2 ± 3.9 HU vs. 33.4 ± 2.3 HU, P < 0.01) and ED values (102.4 ± 0.4 %ED relative to water (%EDW) [ED relative to water] vs. 102.9 ± 0.17 %EDW, P < 0.01) but not in effective Z values (7.4 ± 0.04 vs. 7.4 ± 0.04, P = 0.46) between the infarcted and contralateral regions [Figure 1]. Diagnostic performance of the quantitative values derived from the conventional 120-kVp, ED, and effective Z images was evaluated for each type. ROC analysis revealed AUC values of 0.85, 0.90, and 0.62 for the conventional 120 kVp, ED, and effective Z images, respectively. For conventional 120-kVp images, the optimal cutoff value was 30.1 HU, with a diagnostic performance of sensitivity 0.69, specificity of 1.00, positive predictive value (PPV) of 1.00, and negative predictive value (NPV) of 0.77. The optimal cutoff for ED was 102.7 %EDW, with sensitivity of 0.81, specificity of 1.00, PPV of 1.00, and NPV of 0.84. For effective Z imaging, the optimal cutoff was 7.32, sensitivity of 0.23, specificity of 1.00, PPV of 1.00, and NPV of 0.57 [Figure 2].

Close

Figure 1:

Quantitative image analysis. (a) Conventional 120 kVp images, (b) ED images, (c) Effective Z images. Box plots illustrating the results of the quantitative image analysis for conventional 120 kVp (a), ED (b), and effective Z (c) images.

Export to PPT

Close

Figure 2:

Comparison of receiver operating characteristic curves of the quantitative image parameters. The diagnostic accuracy of electron density imaging was superior (area under curve [AUC] = 0.90) compared with that of conventional 120-kVp CT (AUC = 0.85) and effective Z imaging (AUC = 0.62). (CT: Computed tomography).

Export to PPT

Qualitative image analysis

In qualitative analysis, the area-based concordance with MRI for observer 1 demonstrated a kappa value of 0.33 for the conventional 120-kVp images and 0.63 for ED images, whereas for observer 2, kappa value was 0.29 for the conventional 120-kVp images and 0.57 for ED images. In ROC analysis, AUC values for observer 1 were 0.62 (95% confidence interval [CI], 0.57–0.67) for the conventional 120-kVp images and 0.78 (95% CI, 0.74–0.82) for ED images, whereas for observer 2, AUC values were 0.60 (95% CI, 0.55– 0.65) for the conventional 120-kVp images and 0.75 (95% CI, 0.71–0.79) for ED images [Figure 3]. The interobserver agreement between observers 1 and 2 yielded a kappa value of 0.65 for the conventional 120-kVp images and 0.71 for ED images. Figures 4 and 5 present representative cases of AIS.

Close

Figure 3:

Comparison of ROC curves in the qualitative image analysis. (a) area under the curve (AUC) values for observer 1 and (b) observer 2. In ROC analysis, area under the curve (AUC) values for observer 1 (a) were 0.62 (95% confidence interval [CI], 0.57–0.67) for conventional 120-kVp images and 0.78 (95% CI, 0.74–0.82) for ED images, whereas for observer 2 (b), AUC values were 0.60 (95% CI, 0.55–0.65) for the conventional 120-kVp images and 0.75 (95% CI, 0.71–0.79) for ED images.

Export to PPT

Close

Figure 4:

A 69-year-old woman with acute ischemic stroke (AIS) in the right occipital lobe. (a) 120 kVp images, (b) ED images, (c) effective Z images, (d) MRI images. Diffusion-weighted MRI revealed an abnormal signal (arrow) in the right occipital lobe. The lesion was poorly visualized on the conventional 120-kVp image but was clearly identifiable on ED imaging. Effective Z imaging demonstrated only a slight difference between the left and right occipital lobes.

Export to PPT

Close

Figure 5:

A 93-year-old man with acute ischemic stroke (AIS) in the left middle cerebral artery territory. (a) 120 kVp images, (b) ED images, (c) effective Z images, (d) MRI images. Diffusion-weighted MRI revealed an abnormal signal (arrow) in the left middle cerebral artery. The lesion was poorly visualized on the conventional 120-kVp image but was clearly identifiable on ED imaging. Effective Z imaging showed a discernible difference between the left and right sides.

Export to PPT

DISCUSSION

In this study, we found statistically significant differences in CT (P < 0.01) and ED values (P < 0.01), but not in the effective Z values, between the infarcted and contralateral brain regions. ROC analysis demonstrated that the diagnostic performance of ED imaging was superior to that of conventional 120-kVp imaging and effective Z measurements. Qualitative analysis showed that ED images demonstrated better agreement with MRI and higher interobserver consistency, with the kappa values reflecting improved diagnostic accuracy and reliability over the conventional 120-kVp images.

DLCT uses a dual-layer detector design in which lower-energy photons are absorbed by an upper scintillator layer and higher-energy photons by a lower layer, enabling spectral data acquisition in a single scan. ED imaging, derived from DLCT, facilitates quantitative tissue characterization by calculating ED relative to water. The probability of Compton scattering is proportional to the number of electrons within a material, that is, ED. Materials with higher ED provide more opportunities for X-rays to interact with electrons, resulting in frequent Compton scattering. Consequently, tissues or materials with high ED show greater Compton scattering when exposed to X-rays. This principle facilitates ED estimation by quantifying Compton scattering across various tissues or materials. For substances with low atomic numbers, such as water and fat, in which Compton scattering predominates, ED imaging adjusted for this effect may be particularly valuable in NCCT. In particular, ED imaging is expected to enhance the visual differentiation of pathologies, such as organ edema, inflammation, and thrombotic lesions, without the use of contrast agents.

Rodriguez-Granillo et al. investigated the utility of ED imaging obtained through DLCT for identifying thrombi without contrast agents.^[8] Moreover, in this vitro study, thrombus samples from mechanical thrombectomy were scanned using ED imaging, conventional CT, and low-monoenergetic imaging; ED imaging had significantly higher signal-to-noise and contrast-to-noise ratios, providing more accurate thrombus visualization. These findings suggest the potential of ED imaging for thrombus detection in patients who are ineligible for contrast-enhanced CT. Bae and Jeon reported a case of acute pulmonary embolism (PE) detected using ED imaging from an unenhanced dual-energy CT.^[9] In an elderly patient presenting with pneumonia symptoms, conventional NCCT showed no signs of PE. However, retrospective ED imaging revealed multiple high-density areas in the pulmonary arteries that corresponded to thrombus locations confirmed later by contrast-enhanced CT pulmonary angiography. This case highlights the potential of ED imaging for identifying PE in patients in whom contrast-enhanced imaging cannot be performed. As mentioned above, although several reports have highlighted the efficacy of ED imaging for thrombus evaluation, our study is the first to demonstrate the utility of ED imaging in NCCT for AIS. The findings from our study demonstrated that DLCT, specifically using ED imaging, enhanced the detection of EICs in patients with AIS compared with that using conventional CT. By providing additional quantitative insights, DLCT may reduce the reliance on MRI for urgent diagnosis, potentially expediting treatment initiation. This advancement is particularly beneficial for clinical settings in which MRI access is limited, ultimately supporting timely intervention and improved patient outcomes in acute stroke care.

Nagano et al. evaluated the use of ED derived from DLCT for diagnosing metastatic mediastinal lymph nodes in non-small-cell lung cancer.^[10] In a study on 57 patients, ED values were significantly lower in the metastatic nodes than in nonmetastatic nodes. Combining ED with short-axis diameter or [18F]-fluorodeoxyglucose uptake on PET/CT improved the diagnostic accuracy, surpassing that of individual parameters, suggesting the potential for expanding clinical applications of ED imaging to the field of oncology.

In dual-energy CT, effective Z imaging provides valuable clinical insights by enabling material differentiation based on the elemental composition. This capability is particularly useful for distinguishing between tissues with similar densities but different atomic structures, such as for characterizing arterial plaques in vascular imaging^[11-13] or assessing tumor tissue characteristics in oncological evaluations.

Our results indicated that effective Z imaging was not functional for evaluating AIS, probably due to the similarity in atomic numbers between the infarcted and non-infarcted areas. In contrast, ED imaging, which reflects differences in tissue Compton scattering, proved to be more effective for detecting AIS in the NCCT setting.

There were several limitations in this study. First, as a single-center study with a limited sample size, our findings require validation in larger prospective studies to ensure generalizability. Second, this study used a specific dual-energy CT system, DLCT system, and a single scanner; therefore, it is unclear if similar results would be obtained when using other dual-energy systems or different equipment. Third, during the qualitative assessment, we adjusted the window settings as needed when visually interpreting the ED images. Future research should investigate the optimal window settings for ED images to standardize and improve the diagnostic accuracy. Finally, unlike diffusion-weighted MRI, abnormalities detected on ED images are non-specific. Therefore, the interpretation of ED images should be conducted in conjunction with the clinical symptoms and physical examination findings to provide a comprehensive assessment.

CONCLUSION

DLCT demonstrated significant advantages in detecting EICs in AIS, with higher diagnostic accuracy and interobserver agreement obtained with ED imaging than with conventional CT imaging. These findings suggest that ED imaging can improve lesion visualization, potentially enhancing diagnostic reliability in emergency stroke settings.