Quantitative assessment of coronary peri-vascular fat attenuation index using dual-layer spectral CT for differentiating high-risk plaques in coronary artery disease

Abstract

Background

Perivascular adipose tissue (PCAT) inflammation contributes to plaque vulnerability in coronary artery atherosclerotic heart disease (CHD). The fat attenuation index (FAI) has emerged as a non-invasive marker of vascular inflammation. However, its association with plaque characteristics, particularly high-risk plaques (HRPs), remains insufficiently understood. This study evaluated coronary peri-vascular FAI using dual-layer spectral CT (SDCT) and its relationship with plaque calcification and HRP identification.

Methods

A retrospective analysis of 182 patients with CHD and 312 plaques was conducted. FAI was measured on 120 kVp images, virtual monoenergetic images, and effective atomic number (Eff-Z) images across energy levels from 40 keV to 100 keV. The spectral slope (λ_HU) was also calculated. The relationship between FAI and plaque calcification was assessed, and differences in quantitative parameters between high-risk and non-high-risk plaque groups were used to determine optimal cut-off values for differentiating HRPs.

Results

As plaque calcification increased, FAI values generally decreased. The fat attenuation values around HRPs were significantly higher than those around non-HRPs. Additionally, λ_Hu (λ_{40-70 keV}, λ_{40-100 keV}) and Eff-Z values were significantly higher in HRPs compared to non-HRPs (all p < 0.001). Multivariate logistic regression identified FAI at 40 keV, λ_{40-70 keV}, and Eff-Z as independent predictors for HRP diagnosis. The combined model of these parameters had the highest diagnostic efficacy, with an AUC of 0.881, sensitivity of 87.1%, and specificity of 75.0%.

Conclusion

FAI assessed using SDCT demonstrates enhanced discrimination of plaque composition and characteristics compared to conventional CT. The combined model of FAI40keV, λ40–70 keV, and Eff-Z shows the highest discriminative ability for high-risk plaque features, suggesting potential value for non-invasive plaque characterization that warrants prospective validation.

Introduction

Atherosclerotic plaques are the pathological basis of Coronary Atherosclerotic Heart Disease (CHD) [1]. The rupture and subsequent thrombosis of high-risk plaques (HRPs) play a critical role in the onset and progression of acute coronary syndrome (ACS) [2–4]. Therefore, accurately assessing the characteristics of HRPs is of significant clinical value in the risk stratification of cardiovascular disease patients.

Peri-coronary Adipose Tissue (PCAT) is an endocrine-active adipose tissue that interacts with adjacent vascular walls [5]. It regulates cardiovascular biological functions through paracrine signaling and also responds to signals from the vascular wall, altering its phenotype [6]. When PCAT function becomes dysregulated, it can secrete various cytokines and inflammatory mediators, leading to endothelial dysfunction and local inflammation. This accelerates the formation and progression of atherosclerotic plaques, thereby increasing the risk of developing HRPs [7].The Fat Attenuation Index (FAI), which quantifies the attenuation of fat tissue surrounding the coronary arteries, has been proposed as a valuable imaging biomarker for evaluating PCAT and its role in atherosclerosis [8, 9]. Elevated FAI values are associated with higher levels of inflammation and may help in differentiating HRPs from non-high-risk ones.

Coronary Computed Tomography Angiography (CCTA) is a non-invasive imaging technique with high sensitivity for detecting HRPs. However, conventional CT has certain limitations in distinguishing tissues with similar densities. Dual-layer Spectrum Detector CT (SDCT) improves tissue density differentiation and can simultaneously generate conventional CT images and multiple spectral parameter images, including Virtual Mono-energetic Images (VMI), spectral slope (λ_HU), and Effective Atomic Number (Eff-Z) maps [10–12].

Currently, the relationship between PCAT attenuation values and plaque composition, plaque characteristics, and the degree of coronary artery stenosis remains unclear. Additionally, considering the localized inflammatory effects within the coronary artery, measuring PCAT attenuation values around plaques may provide more accurate diagnostic information. Therefore, this study aims to explore the correlation between spectral parameters, such as VMI, λ_Hu, and Eff-Z, derived from SDCT imaging and plaque calcification ratios. The study also seeks to identify the most suitable imaging parameters for detecting inflammation around HRPs.

Method

Study population

This retrospective study consecutively enrolled 295 patients with known or suspected CHD who underwent SDCT at the Second Affiliated Hospital of Nanchang University between December 2021 and December 2023. After applying the inclusion and exclusion criteria, 113 patients were excluded, resulting in a final cohort of 182 patients and 312 coronary atherosclerotic plaques included for analysis.

This study was approved by the Ethics Committee of the Second Affiliated Hospital of Nanchang University , and the requirement for informed consent was waived due to its retrospective design.

The exclusion criteria were as follows: [1] diffuse coronary artery disease without discrete plaque formation on CT imaging; [2] plaques confined to small branches of the left main coronary artery or other distal coronary branches; [3] congenital or acquired anatomic variations of the heart or coronary arteries; [4] history of myocardial infarction or cardiac surgery; [5] inadequate peri-coronary adipose tissue (PCAT) or poor image quality that precluded reliable post-processing analysis.

Baseline clinical characteristics were collected for all enrolled patients, including age, sex, history of hypertension, hyperlipidemia, diabetes mellitus, smoking, alcohol consumption, and laboratory findings. Hypertension was defined according to the 2024 European Society of Cardiology (ESC) guidelines [13]. Baseline clinical variables included smoking and alcohol consumption status, both of which were defined as active use at the time of CCTA or discontinuation within 6 months prior to imaging. Diabetes mellitus and hyperlipidemia were defined based on clinical diagnoses established by the treating physicians [14].

Patient stratification

To address different research aims, enrolled patients and plaques were stratified accordingly:

Analysis based on plaque calcification extent

Previous studies have indicated that PCAT surrounding HRP may exhibit inflammatory changes, potentially affecting the interpretation of PCAT attenuation values [15]. To minimize this confounding effect and focus on the association between PCAT attenuation and intraplaque calcification, plaques identified as HRP were excluded.

The remaining plaques were classified as non-calcified (Group A), mixed (Group B), or calcified (Group C) in accordance with the guidelines of the Society of Cardiovascular Computed Tomography (SCCT) [16, 17]. Plaque composition was classified based on voxel attenuation thresholds derived through automated volumetric analysis. Specifically, voxels with attenuation > 130 HU were defined as calcified tissue. The calcification ratio was calculated as the percentage of calcified volume divided by the total plaque volume. Consistent with criteria used in previous quantitative CCTA studies, plaques were categorized as non-calcified (Group A) when the calcification ratio was < 10%, mixed (Group B) when 10% ≤ calcification ratio < 90%, and calcified (Group C) when the calcification ratio was ≥ 90% [18, 19]. Quantitative parameters derived from SDCT, including VMI,λ_HU, and Eff-Z, were compared across the three groups to explore their correlation with plaque calcification extent.

Analysis for HRP identification

HRPs were defined by the presence of at least two of the following four established CT high-risk features: low-attenuation plaque, positive remodeling, spotty calcification, and the napkin-ring sign [20]. Given that HRPs are typically characterized by low attenuation and lipid-rich necrotic cores, plaques composed exclusively or predominantly of calcification were excluded to enhance the specificity of parameter selection for detecting HRP-associated inflammation. Quantitative PCAT parameters were compared between HRP and non-HRP plaques. To clarify the source of non-high-risk plaques, they were further categorized into two subgroups: (1) Non-HRP①, defined as non-high-risk plaques identified in patients who also harbored at least one HRP (HRP-positive group); and (2) Non-HRP②, defined as non-high-risk plaques identified in patients with no HRPs (Control group).

A schematic overview of the study workflow is provided in Fig. 1.

Fig. 1.

Study workflow and group comparisons. Abbreviations: CHD = coronary heart disease; SDCT = spectral detector CT; HRP = high-risk plaque; MI = myocardial infarction

CT scanning protocol

All CT examinations were performed on a dual-layer spectral detector CT scanner (IQon Spectral CT, Philips Healthcare, Best, the Netherlands). A non-enhanced scan was routinely performed prior to contrast-enhanced SDCT acquisition to calculate the coronary artery calcium (CAC) score using the Agatston method.

All contrast-enhanced scans were performed using a standardized protocol to ensure consistency in arterial opacification and reduce variability in perivascular attenuation measurement. An 18-gauge intravenous catheter was placed in the antecubital vein, and iodixanol (370 mg I/mL, Bayer Healthcare) was administered using a dual-head high-pressure injector (Ulrich Medical, Germany).

The total volume of contrast agent was calculated according to body weight (0.8 mL/kg), and the injection rate was individualized (typically 4–6 mL/s) based on the calculated volume and a target injection duration of approximately 14 s. To maintain consistent enhancement timing, bolus tracking was employed in all patients. The region of interest (ROI) was placed in the ascending aorta at the level of the pulmonary artery bifurcation. Scanning was automatically triggered with a 6-second delay once the ROI attenuation reached 120 Hounsfield units.

A 30–40 mL saline flush was administered immediately after the contrast bolus at the same injection rate to optimize contrast utilization and reduce streak artifacts.

Image analysis

Plaque morphology and composition analysis

Plaque characteristics, including luminal stenosis, length, volume, remodeling index, and composition, were automatically quantified using the semi-automated CCTA/SDCT software CoronaryDoc^® (ShuKun Network Technology, Beijing, China). Stenosis and remodeling were assessed based on proximal and distal reference diameters, while plaque length and volume were calculated along the vessel centerline.

Peri-plaque adipose tissue analysis

Quantitative assessment of PCAT surrounding coronary plaques was performed using the same platform on conventional CT images (120 kVp) as well as virtual monoenergetic images (VMI) ranging from 40 to 100 keV (in 10-keV intervals). The FAI was measured along a cylindrical volume of interest (VOI) centered on each target plaque (volume ≥ 2 mm^³) and aligned with its longitudinal axis. Corresponding FAI values were recorded as FAI₁₂₀kVp, and FAI₄₀keV through FAI₁₀₀keV for subsequent analysis.

Since CT attenuation is inherently energy-dependent, PCAT attenuation has been shown to vary systematically with both tube voltage and monoenergetic energy levels [9, 21, 22]. Consequently, the conventional adipose tissue window defined for 120-kVp images (–190 to − 30 HU) cannot be directly applied to low-keV VMI, particularly at 40–70 keV [21]. Chen et al. validated energy-specific thresholds for 40 keV (–280 to − 40 HU) and 70 keV (–220 to − 30 HU). For the remaining energy levels, thresholds were derived by linear interpolation between these two anchor points and extrapolation toward the standard 120 kVp window, based on the approximately linear relationship between monoenergetic energy level and adipose tissue attenuation [9, 23]. This interpolation approach and the resulting threshold series have been independently adopted in a recent large-scale SDCT study [24]. The resulting energy-specific voxel density thresholds were: [–280, − 40 HU], [–260, − 40 HU], [–240, − 40 HU], [–220, − 30 HU], [–210, − 30 HU], [–200, − 30 HU], and [–190, − 30 HU] for images at 40–100 keV (in 10-keV intervals), respectively, while − 190 to − 30 HU was used as the standard fat attenuation range on conventional 120 kVp images.

Effective atomic number (Eff-Z) measurement

CCTA datasets were post-processed using IntelliSpace Portal software (ISP version 10.1.5, Philips Healthcare, the Netherlands), the vendor-recommended platform for spectral analysis, to generate effective atomic number (Eff-Z) maps. Eff-Z measurements were performed on curved planar reformatted (CPR) images. For each lesion, ROIs were placed in the peri-plaque adipose tissue surrounding the target lesion. Three to five CPR slices at varying angulations were selected to capture representative peri-plaque adipose tissue regions, and the average Eff-Z value across these slices was computed and used for subsequent analysis. This dual-platform approach (CoronaryDoc^® for FAI; ISP for Eff-Z) is consistent with prior SDCT studies [24–26] (Fig. 2).

Fig. 2.

Schematic diagram of effective atomic number (Eff-Z) measurement of peri-plaque adipose tissue on curved planar reformatted images using IntelliSpace Portal (Philips Healthcare)

Spectral attenuation features

In spectral curve images, CT values generally increase with energy, but the rate of increase diminishes as energy rises. To highlight the spectral attenuation characteristics of adipose tissue, energy ranges of 40–70 keV, 40–100 keV, and 70–100 keV were selected to calculate the spectral curve slope (λ) for adipose tissue. The specific calculation method is as follows:

The images were analyzed and processed by two radiologists. All automated measurements were independently reviewed by experienced readers, and manual adjustments were made when necessary to ensure accuracy and reproducibility.

Statistical analysis

Data analyses were performed using IBM SPSS Statistics version 27.0, with statistical graphs generated via GraphPad Prism 9.0 and MedCalc 15.8 to visualize data distribution and comparison results. The Intraclass correlation coefficient (ICC) was used to assess measurement repeatability.

Normality was assessed using the Shapiro-Wilk test. Normally distributed continuous data were expressed as mean ± standard deviation, while skewed data were reported as median (interquartile range). Categorical data were presented as percentages.

For group comparisons, appropriate statistical methods were chosen based on data type and distribution. Independent samples t-tests or Mann-Whitney U tests were used for continuous data between two groups, and chi-square tests were applied for categorical data. One-way ANOVA (with LSD post-hoc test) or Kruskal-Wallis test (with Bonferroni correction) was used for comparisons among three groups, with chi-square tests (and Bonferroni correction) applied for categorical data.

Pearson or Spearman correlation coefficients were used to assess associations between continuous variables. Binary multivariate logistic regression evaluated HRP risk factors and their incremental value. Receiver operating characteristic (ROC) curves were plotted to determine optimal cutoff values, sensitivity, and specificity, with area under the curve (AUC) calculations. Factors influencing PCAT attenuation were analyzed using multivariate linear regression, with PCAT attenuation as the dependent variable.

A p-value < 0.05 was considered statistically significant for all analyses.

Reliability analysis

Using the ICC as the primary metric, 60 plaques were randomly selected for repeat post-processing after a two-week interval to assess intra-observer agreement. Inter-observer agreement was evaluated by comparing measurements between the two independent readers.

Results

Repeatability analysis

The FAI based on multiparametric measurements using SDCT demonstrate good consistency in intra-observer and inter-observer repeatability analyses. The ICC values and 95% confidence intervals are detailed in Table 1.

Table 1.

Reproducibility analysis of spectral parameters intra- and inter-observer

Variable	Inter-observer		Intra-observer
	ICC	95%CI	ICC	95%CI
FAI40 keV	0.824	(0.692, 0.898)	0.866	(0.785, 0.918)
FAI50 keV	0.828	(0.573, 0.918)	0.860	(0.776, 0.914)
FAI60 keV	0.912	(0.850, 0.948)	0.939	(0.897, 0.963)
FAI70 keV	0.881	(0.809, 0.927)	0.900	(0.838, 0.939)
FAI80 keV	0.874	(0.797, 0.922)	0.932	(0.889, 0.959)
FAI90 keV	0.853	(0.766, 0.910)	0.939	(0.900, 0.963)
FAI100 keV	0.831	(0.732, 0.895)	0.930	(0.885, 0.957)
Eff-Z	0.895	(0.830, 0.936)	0.936	(0.895, 0.961)

Note. FAI Fat Attenuation Index, Eff-Z Effective Atomic Number. The unit of FAI in the figure is HU

Clinical baseline characteristics

A total of 182 patients with 312 plaques were included in this study. Among them, 58 patients with at least one HRP were assigned to the HRP-positive group, while 69 patients without HRP constituted the control group. The clinical baseline characteristics of all patients are presented in Table 2. There were no statistically significant differences between the HRP-positive and control groups in terms of age, gender distribution, body mass index (BMI), or cardiovascular-related clinical symptoms (including chest tightness, palpitations, dyspnea, syncope, and chest pain).

Table 2.

Comparison of clinical data between the HRP-positive group and the control group

Variable	Total (N = 182)	HRP-positive Group (N = 58)	Control Group (N = 69)	p
Age, y	61.30 ± 11.47	62.60 ± 12.72	61.64 ± 10.05	0.641
Male	118(64.84)	42(72.41)	45(65.22)	0.384
BMI, Kg/m²	24.79 ± 3.02	24.68 ± 2.73	25.19 ± 2.93	0.311
Chest tightness	141(77.47)	44(75.86)	56(81.16)	0.467
Dyspnea	66(36.26)	19(32.76)	25(36.23)	0.682
Syncope	9(4.95)	5(8.62)	4(5.80)	0.787
Chest pain	62(34.07)	18(31.03)	25(36.23)	0.538
Palpitations	40(21.98)	6(10.34)	14(20.29)	0.125
Smoking	64(35.16)	19(32.76)	31(44.93)	0.162
Alcohol history	52(28.57)	13(22.41)	24(34.78)	0.126
Hypertension	97(53.30)	33(56.90)	45(65.22)	0.337
Hyperlipidemia	80(43.96)	31(53.45)	23(33.33)	0.022
Diabetes	79(43.41)	30(51.72)	29(42.03)	0.275
TC	4.72 ± 1.13	4.95 ± 1.07	4.65 ± 0.97	0.103
TG	1.44(1.02, 2.04)	1.69(1.05, 2.34)	1.21(0.99, 1.98)	0.042
HDL	1.27 ± 0.30	1.30 ± 0.33	1.28 ± 0.26	0.634
LDL	2.49 ± 0.78	2.72 ± 0.92	2.41 ± 0.68	0.030
Hcy	12.89(10.98, 15.48)	12.78(11.26, 16.81)	13.37(10.95, 15.47)	0.735
Statins	55(30.22)	17(29.31)	20(28.99)	0.968

Note: p < 0.05, HRP-positive Group vs. Control Group. BMI: Body Mass Index; TC: Total Cholesterol; TG: Triglycerides; HDL: High-Density Lipoprotein; LDL: Low-Density Lipoprotein; Hcy: Homocysteine

Regarding cardiovascular risk factors and laboratory parameters, the HRP-positive group had a significantly higher prevalence of hyperlipidemia and elevated low-density lipoprotein (LDL) levels compared to the control group, with both differences being statistically significant (p < 0.05).

Relationship between FAI and plaque calcification characteristic

Tables 3 and 4 summarize the anatomical characteristics and FAI values of plaques stratified by three levels of calcification. Calcified plaques show significantly more lumen narrowing than non-calcified plaques (p < 0.001), while mixed plaques are longer than non-calcified plaques (p < 0.05). No significant differences were observed in lesion location or plaque volume, nor in the remodeling index, a key HRP risk factor, among the groups.

Table 3.

Comparison of anatomical characteristics of plaques with different calcification ratios

Variable	Group A (n = 73)	Group B (n = 51)	Group C (n = 126)	p
Plaque distribution location				0.980
LAD, n	36(49.32)	27(52.94)	65(51.59)
LCX, n	10(13.7)	8(15.69)	18(14.29)
RCA, n	27(36.99)	16(31.37)	43(34.13)
Degree of stenosis, %	33.96 ± 11.70	37.43 ± 16.90	42.76 ± 12.82	<.001a
Plaque length, mm	13.65 ± 6.28	16.41 ± 6.87	15.26 ± 5.28	0.033b
Plaque volume, mm³	41.13(23.68, 56.05)	47.94(35.00, 65.68)	41.59(29.91, 65.67)	0.071
Remodeling index	1.12 ± 0.11	1.01 ± 0.11	0.84 ± 0.11	0.100

Note: a p < 0.001, Group A vs. Group C; b p < 0.05, Group A vs. Group B. LAD: Left Anterior Descending artery; LCX: Left Circumflex artery; RCA: Right Coronary artery

Table 4.

Comparison of FAI values of plaques with different calcification ratios

Variable	Group A (n = 73)	Group B (n = 51)		Group C (n = 126)	p	p1	p2	p3
FAI120kVp	-76.6 ± 6.5	-79.8 ± 8.7		-83.5 ± 6.4	<0.001	0.088	<0.001	0.020
FAI40keV	-131.3 ± 14.0	-146.1 ± 15.0		-153.7 ± 12.4	<0.001	<0.001	<0.001	<0.001
FAI50keV	-113.8 ± 11.7	-120.4 ± 11.7		-124.2 ± 9.8	<0.001	<0.001	<0.001	0.035
FAI60keV	-97.3 ± 12.4	-101.4 ± 10.6		-105.1 ± 10.9	<0.001	0.047	<0.001	0.048
FAI70keV	-80.8 ± 8.4	-84.0 ± 7.7		-87.1 ± 6.7	<0.001	0.020	<0.001	0.012
FAI80keV	-76.8 ± 9.0	-78.5 ± 6.8		-79.7 ± 5.8	0.038	0.568	0.040	0.555
FAI90keV	-73.5 ± 7.7	-74.5 ± 5.8		-75.8 ± 6.0	0.083	0.794	0.099	0.486
FAI100keV	-71.4 ± 7.1	-71.9 ± 5.9		-73.4 ± 5.5	0.070	0.968	0.118	0.309
λ40–70 keV	-1.68 ± 0.40	-2.07 ± 0.48		-2.22 ± 0.32	<0.001	<0.001	<0.001	0.121
λ40–100 keV	-1.00 ± 0.22	-1.24 ± 0.25		-1.34 ± 0.18	<0.001	<0.001	<0.001	0.003
λ70–100 keV	-0.30 (-0.43, -0.13)	-0.37 (-0.60, -0.20)		-0.43 (-0.57, -0.33)	<0.001	0.095	<0.001	0.300
Eff-Z	6.45 (6.27, 6.62)	6.28 (6.03, 6.40)		6.32 (6.06, 6.58)	0.005	0.005	0.055	0.530

Note: p-value: Overall statistical difference among Groups A, B, and C; p1-value: Statistical difference between Group A and Group B; p2-value: Statistical difference between Group A and Group C; p3-value: Statistical difference between Group B and Group C. The unit of FAI in the figure is HU

FAI generally decreases from Group A to Group C. FAI values measured at 40 keV showed significant differences across all pairwise comparisons (-131.34 ± 14.03 vs. -146.08 ± 15.01 vs. -153.72 ± 12.36, all p < 0.001), with similar results for 50–70 keV measurements. However, for higher energy levels, a significant difference was only found between Groups A and C at 80 keV (Fig. 3A).

Fig. 3.

Comparison of quantitative spectral parameters among plaques with different calcification degrees (Group A: non-calcified; Group B: mixed; Group C: calcified). The panels illustrate differences in (A) Fat Attenuation Index (FAI) derived from virtual monoenergetic images (VMI), (B) spectral attenuation curve slope (λ), and (C) effective atomic number (Eff-Z). “◇” indicates p < 0.05 and “☆” indicates p < 0.001

The spectral shapes of the three plaque groups were similar (Fig. 3B). The spectra of mixed and calcified plaques (red and yellow curves) nearly overlapped, both below the non-calcified plaque spectrum (blue curve). As energy levels increased, group differences decreased. The slopes of the spectral curves (λ_40−70keV and λ_40−100keV) showed significant differences between Groups A and B, and A and C (all p < 0.05; Fig. 3C). Eff-Z values were significantly different between Groups A and B (p < 0.05). Spearman’s correlation analysis revealed a strong negative correlation between calcified component proportion and FAI at 120 kVp, 40 keV, λ_40−70keV, and λ_40−100keV, with correlation coefficients (r) of -0.601, -0.578, -0.502, and − 0.554, respectively (all p < 0.001).

Relationship between FAI and plaque vulnerability

Patients were divided into the HRP-positive group and control group based on the presence of HRPs. The HRP-positive group included 58 patients with 62 high-risk and 35 non-high-risk plaques, while the control group consisted of 69 patients with 89 non-HRPs. A comparison of plaque characteristics and FAI across the three groups is presented in Tables 5 and 6.

Table 5.

Comparison of plaque characteristics between the control and the HRP-positive group

Variable	HRP-positive Group (N = 58)		p1	Control Group (N = 69)	p2
	HRPs(n = 62)	Non-HRPs①(n = 35)		Non-HRPs②(n = 89)
Plaque distribution location			0.730		0.907
LAD	37(59.68)	18(51.43)		53(59.55)
LCX	7(11.29)	5(14.29)		12(13.48)
RCA	18(29.03)	12(34.29)		24(26.97)
HRP characteristics			0.015		0.025
spotty calcification	24(38.71)	5(14.29)		6(6.74)
low-density attenuation	53(85.48)	1(2.90)		6(6.74)
napkin-ring sign	17(27.42)	1(2.90)		0
positive remodeling	52(83.87)	11(31.43)		18(20.22)
Degree of stenosis, %	37.63 ± 15.36	34.34 ± 13.88	0.298	36.71 ± 12.48	0.686
Plaque length, mm	14.95 ± 5.07	16.77 ± 8.04	0.232	14.57 ± 6.40	0.690
Plaque volume, mm³	39.54 (24.87, 55.51)	43.68 (28.00, 70.46)	0.217	45.32 (28.44, 64.84)	0.258
Remodeling index	1.19 ± 0.12	1.09 ± 0.06	<0.001	1.07 ± 0.09	<0.001
Lipid content, mm³	10.59 (4.01, 15.17)	3.03 (1.08, 9.72)	<0.001	4.57 (1.75, 10.14)	<0.001
Fibrous lipid content, mm³	25.26 (15.43, 40.21)	10.97 (4.82, 26.75)	<0.001	16.2 (9.43, 24.37)	<0.001
Fibrous content, mm³	16.22 (10.45, 27.69)	15.45 (6.06, 21.19)	0.247	9.89 (4.97, 17.95)	<0.001
Calcified content, mm³	0.67 (0, 10.37)	7.55 (2.24, 16.92)	0.002	0 (0, 3.70)	0.047
Lipid content percentage, %	15.75 (9.12, 22.93)	6.43 (2.96, 17.08)	<0.001	13.9 (7.42, 19.82)	0.260
Fibrous lipid percentage, %	45.79 ± 14.16	32.31 ± 14.13	<0.001	47.03 ± 15.77	0.621
Fibrous content percentage, %	29.61 ± 12.60	34.10 ± 12.49	0.095	29.59 ± 12.59	0.992
Calcified content percentage, %	0.97 (0, 11.85)	24.45 (3.7, 41.79)	<0.001	0 (0, 12.38)	0.191

Note: p1-value: Statistical difference between HRP and Non-HRP plaques within the HRP-positive Group; p2-value: Statistical difference between HRP plaques in the HRP-positive Group and Non-HRP plaques in the control group

Table 6.

Comparison of FAI values of plaques between the control and HRP-positive group

Variable	HRP-positive Group (N = 58)		p1	Control Group (N = 69)	p2
	HRPs (n = 62)	Non-HRPs① (n = 35)		Non-HRPs② (n = 89)
FAI120kVp	-77.1 ± 8.6	-82.5 ± 7.9	0.003	-80.0 ± 8.6	0.038
FAI40keV	-127.2 ± 13.4	-143.3 ± 16.7	<0.001	-146.8 ± 16.6	<0.001
FAI50keV	-109.9 ± 11.5	-119.3 ± 12.9	<0.001	-121.5 ± 11.1	<0.001
FAI60keV	-94.5 ± 10.0	-102.6 ± 14.6	0.005	-103.8 ± 10.8	<0.001
FAI70keV	-83.0 ± 10.8	-87.7 ± 10.8	0.045	-89.1 ± 13.2	0.003
FAI80keV	-74.5 ± 8.3	-78.1 ± 7.9	0.037	-79.7 ± 8.0	<0.001
FAI90keV	-73.6 ± 7.1	-71.5 ± 7.7	0.178	-75.2 ± 7.1	0.003
FAI100keV	-71.9 ± 6.8	-70.0 ± 7.1	0.193	-72.3 ± 6.7	0.044
λ40–70 keV	-1.47 ± 0.30	-1.86 ± 0.44	<0.001	-1.92 ± 0.40	<0.001
λ40–100 keV	-0.95 ± 0.19	-1.19 ± 0.23	<0.001	-1.24 ± 0.26	<0.001
λ70–100 keV	-0.47 (-0.70, -0.30)	-0.43 (-0.53, -0.27)	0.136	-0.47 (-0.70, -0.30)	0.091
Eff-Z	6.59 ± 0.27	6.28 ± 0.30	<0.001	6.26 ± 0.29	<0.001

Note: p1-value: Statistical difference between HRP and Non-HRP plaques within the HRP-positive Group; p2-value: Statistical difference between HRP plaques in HRP-positive Group and Non-HRP plaques in the control group. The unit of FAI in the figure is HU

At the vessel level, no significant differences in plaque distribution were observed between the HRP-positive and control groups. Among the four HRP features (spotty calcification, positive remodeling, “napkin-ring” sign, and low-attenuation plaques), the HRPs had prevalence rates of 38.71%, 83.87%, 27.42%, and 85.48%, respectively. No significant differences in lumen narrowing, plaque length, or volume were found between non-HRP plaques within or between groups.

HRPs had significantly higher FAI compared to non-HRPs within both groups, regardless of whether measured using 120kVp images or VMI (all p < 0.05). The differences were particularly marked at lower energy levels, with substantial increases in FAI at 40 keV and 50 keV for HRPs (p < 0.001). Representative FAI for the HRP-positive and control groups are shown in Fig. 4A.

Fig. 4.

Comparison of quantitative spectral parameters among the HRP-positive and Control groups. The panels illustrate differences in (A) Fat Attenuation Index (FAI) derived from virtual monoenergetic images (VMI), (B) spectral attenuation curve slope (λ), and (C) effective atomic number (Eff-Z). “◇” indicates p < 0.05 and “☆” indicates p < 0.001

As shown in Fig. 4B, the spectral curve slopes for the three plaque groups differed, with the HRPs having significantly higher λ_Hu values (λ_40-70keV and λ_40-100keV) compared to the non-HRPs (both p < 0.001). Similarly, the Eff-Z of the fat surrounding HRP plaques showed the same significant results (p < 0.001; Fig. 4C).

Correlation analysis revealed that the volume (r = 0.307, p < 0.001) and percentage (r = 0.454, p < 0.001) of fatty components were positively correlated with lesion-specific FAI values.

Evaluation of FAI in diagnosing HRP

In the 127 patients, 62 HRPs and 124 non-HRPs were identified. Variables with statistical significance in univariate logistic regression, including FAI at 120 kVp, FAI at VMI, λ40-70 keV, and Eff-Z values, were used as independent variables in binary logistic regression with HRP presence as the dependent variable. The results showed that FAI at 40 keV, λ_40-70 keV, and Eff-Z values provided incremental diagnostic value for HRP (all p < 0.01, OR > 1, Table 7).

Table 7.

Logistic regression analysis of the relationship between spectral parameters and HRP

Variable	B	p	Univariate OR (95% CI)	B	p	Multivariate OR (95% CI)
FAI120kVp	0.051	0.007	1.05(1.01, 1.09)
FAI40 keV	0.078	<0.001*	1.08(1.05, 1.11)	0.056	0.002#	1.06 (1.02, 1.10)
FAI50 keV	0.078	<0.001*	1.08(1.05, 1.11)
FAI60 keV	0.070	<0.001*	1.07(1.04, 1.11)
FAI70 keV	0.042	0.004#	1.04(1.01, 1.07)
FAI80 keV	0.074	<0.001*	1.08(1.03, 1.12)
FAI90 keV	0.062	0.006#	1.06(1.02, 1.11)
FAI100 keV	0.047	0.041#	1.05(1.00, 1.10)
λ40–70 keV	3.521	<0.001*	33.82(10.09, 113.35)	1.899	0.019#	6.68 (1.36, 32.78)
Eff-Z	3.916	<0.001*	50.22(13.14, 191.98)	3.549	<0.001*	34.77 (7.86, 153.80)

Note: “*” and “#” indicate statistically significant differences; “*” denotes p < 0.001; “#” denotes p < 0.05. The unit of FAI in the figure is HU

ROC curve analysis was conducted to assess the diagnostic performance of FAI at 40 keV, λ40–70 keV, and Eff-Z, as well as a combined regression model. The corresponding AUCs were 0.802, 0.796, 0.792, and 0.881, respectively. Notably, the combined model demonstrated superior diagnostic accuracy, with a sensitivity of 87.1% and a specificity of 75.0%. Detailed results are presented in Table 8 and illustrated in Fig. 5.

Table 8.

Diagnostic performance of spectral CT FAI for HRPs

Variable	AUC	Cut-off	95%CI	Sensitivity (%)	Specificity (%)
FAI120kVp	0.634	-83.5	(0.551, 0.717)	82.3	41.9
FAI40keV	0.802	-135.5	(0.739, 0.864)	74.2	72.6
λ40-70 keV	0.796	-1.82	(0.732, 0.861)	87.1	56.5
Eff-Z	0.792	6.41	(0.724, 0.860)	82.3	70.2
Combined model	0.881	0.31	(0.833, 0.929)	87.1	75.0

Note: AUC: Area Under the Curve; Cut-off: Optimal cutoff value

Fig. 5.

Receiver operating characteristic (ROC) curve analysis of SDCT-based FAI parameters for diagnosing HRP. The combined model included FAI40keV, λ40–70 keV, and Eff-Z. This model achieved the highest area under the curve (AUC)

Discussion

This study used SDCT to assess plaque anatomy and FAI, revealing two key findings: (1) A strong correlation between plaque calcification ratio and FAI, particularly at 120 kVp, 40 keV, and λ_40-70keV (correlation coefficients: -0.601, -0.578, -0.502, respectively); and (2) The FAI around HRPs was significantly higher than around non-HRPs, with VMI at 40 keV outperforming 120 kVp in HRP evaluation. The combined model of VMI at 40 keV, λ_HU, and Eff-Z achieved the highest diagnostic performance (AUC = 0.881).

Our findings show that as the proportion of calcified components in plaques increases, FAI values decrease, especially at 40 keV, reflecting a decline in FAI as coronary atherosclerosis progresses. This is consistent with Ma et al., who found higher PCAT attenuation values in non-calcified and mixed plaques compared to calcified ones, and in mild stenosis compared to severe stenosis [27]. However, other studies suggest that increased plaque calcification may elevate PCAT attenuation values, possibly due to low-grade inflammation promoting fat tissue remodeling [28–30]. These discrepancies highlight the dynamic nature of plaque development. In our study, non-calcified plaques were likely associated with more inflammatory fat tissue, contributing to a higher water/fat ratio near the vessel wall [6, 31]. As plaques stabilize, with an increase in fibrous and calcified components, PCAT attenuation values decrease [32]. To minimize beam-hardening artifacts caused by dense calcification, we excluded plaques predominantly or entirely calcified. While this exclusion ensures the reliability of our FAI measurements, we acknowledge that minor calcification in mixed plaques may still introduce residual confounding effects.

This study confirms that lower-energy spectral CT enhances the differentiation of coronary artery surrounding fat. However, no significant differences in FAI values were observed at 90 keV and 100 keV, suggesting that VMI above 90 keV may not be optimal for FAI analysis. Larger sample sizes are needed to validate these findings.

At the plaque level, we found that the FAI of patients with HRPs was significantly higher than that of non-HRP patients, whether measured from 120 kVp or VMI images, both within-group and compared to the control group. As the energy level increased, the difference between HRPs and non-HRPs diminished [33]. This finding aligns with previous studies. Yuvaraj et al. performed a matching analysis comparing 41 stable CHD patients with HRPs to 41 without HRPs and found significantly higher RCA-PCAT values in the HRP group. The study also observed higher PCAT attenuation values in patients who experienced ACS compared to those who did not [34].

Previous studies have demonstrated that PCAT attenuation is significantly elevated around high-risk plaques, with low-energy monoenergetic images—particularly at 40 keV—consistently outperforming conventional 120-kVp imaging in identifying adverse plaque features [21, 25, 26, 35]. Our findings are consistent with this evidence: FAI measured at 40 keV showed superior discrimination between HRPs and non-HRPs compared with higher-energy VMI or standard 120-kVp images. Moreover, we extended prior work by demonstrating that a combined spectral model incorporating FAI_40keV, λ_40–70keV, and Eff-Z provided the highest diagnostic accuracy for HRP identification (AUC = 0.881; sensitivity 87.1%; specificity 75.0%). To ensure high diagnostic accuracy, we excluded plaques in small branches or associated with diffuse disease. While this limits the generalizability to small-vessel disease, it was crucial to minimize partial volume effects and maintain spectral analysis precision in major coronary arteries.

PCAT attenuation reflects a combination of local vascular biology and systemic physiological states (e.g., inflammation, metabolic disorders). Despite baseline lipid heterogeneity, the significantly higher FAI around HRPs compared to non-HRPs within the same patient provides strong internal validation, indicating that spectral changes are not merely due to systemic confounding. Future studies with multivariable adjustments are needed to better assess the impact of these factors, enhancing plaque vulnerability assessment accuracy.

Beyond attenuation-based metrics, our study also demonstrated distinct differences in spectral curves and λ values across plaque types, suggesting their potential role as complementary indicators of plaque vulnerability. In contrast, FAI_100keV and λ_70–100keV showed no significant differences between HRP and non-HRP plaques, likely due to the reduced contrast and lower attenuation variability at higher monoenergetic levels.

Although research on the application of Eff-Z in cardiovascular diseases is still limited, Zhu et al. demonstrated significant differences in FAI_40keV, λ_HU, and Eff-Z between patients with severe coronary artery stenosis and those with normal or non-significant stenosis [26]. Our study further confirms the clinical value of Eff-Z imaging in differentiating plaque types and diagnosing HRPs, underscoring its feasibility for cardiovascular imaging research. Mechanistically, vascular inflammation inhibits adipogenesis and promotes lipolysis, leading to local edema and neovascularization [6]. Since these water- and protein-rich components possess a higher effective atomic number than triglycerides, their accumulation directly elevates Eff-Z values in inflamed PCAT [36, 37]. Recent research has confirmed the technical robustness of SDCT for adipose tissue quantification. A combined phantom and in vivo study showed SDCT achieves high accuracy in fat quantification (mean difference < 1% vs. MRI) and excellent reproducibility (ICC > 0.90), highlighting its methodological consistency [38]. Against this background, further work focusing on spectral slope and Eff-Z imaging for plaque characterization is likely to represent an important future direction in cardiovascular imaging research.

Limitation

This study has several limitations. First, being a retrospective analysis with a small sample size, it may introduce selection bias and limit causal inference. Second, the absence of longitudinal follow-up data prevents evaluating the long-term prognostic value of spectral parameters on clinical outcomes, which requires validation in future prospective studies. Third, HRP identification relied on surrogate markers from CCTA guidelines, not histopathology or intravascular imaging (e.g., OCT or IVUS), limiting plaque characterization precision. Fourth, the energy-specific PCAT attenuation thresholds for intermediate keV levels were derived by linear interpolation from two validated anchor points (40 and 70 keV), and direct verification through VOI volume comparison across energy levels was not performed due to software limitations. Finally, reliance on specific dual-layer spectral CT hardware restricts the method’s widespread applicability. However, this study provides proof-of-concept for the diagnostic value of these spectral parameters. Future research should incorporate multimodal validation and advanced technologies, such as AI, to improve the accuracy and clinical utility of PCAT-based plaque assessment [39, 40].

Conclusion

In conclusion, this study demonstrates a significant correlation between plaque calcification ratio and spectral PCAT parameters. The combined spectral model incorporating FAI at 40 keV, λ40–70 keV, and Eff-Z showed the highest discriminative ability for high-risk plaque features among the parameters evaluated, suggesting that SDCT-derived spectral markers may serve as promising complementary tools for non-invasive plaque characterization. Future prospective studies with invasive imaging validation are warranted to further establish their clinical utility.

Abbreviations

CHD

Coronary Atherosclerotic Heart Disease

ACS

Acute Coronary Syndrome

MACE

Major Adverse Cardiovascular Events

HRP

High-Risk Plaque

PCAT

Peri-coronary Adipose Tissue

CCTA

Coronary CT Angiography

SDCT

Dual-layer Spectrum Detector CT

LAD

Left Anterior Descending branch

LCX

Left Circumflex branch

RCA

Right Coronary Artery

CPR

Curved Planar Reformation

BMI

Body Mass Index

FAI

Fat Attenuation Index

HDL

High-density Lipoprotein

VMI

Virtual Monoenergetic Images

TG

Triglyceride

LDL

Low-density Lipoprotein

Hcy

Homocysteine

ICC

Intraclass Correlation Coefficient

ROC

Receiver Operating Characterisitic curve

AUC

Area Under Curve

Eff-Z

Effective Z

λ_HU

Spectral attenuation curve

Author contributions

Yipei Song contributed to study design, statistical analysis, and manuscript preparation.Ao Kan co-designed the study and participated in data analysis.Ziyan Feng and Mengni Jin performed the literature review.Ying Liu contributed to statistical analysis and conducted clinical data collection.Xiwen Wang assisted with clinical case inclusion.Mengyao Hu participated in experimental studies and data analysis.Jingjing Zhou and Qiming Fang helped with manuscript drafting and editing.Lianggeng Gong supervised the project and served as the guarantor of the study’s integrity.All authors read and approved the final manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (Grant No. 82260342).

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to patient confidentiality and institutional regulations. However, key anonymized data supporting the main findings of this study are available from the corresponding author upon reasonable request and with approval from the Ethics Committee of the Second Affiliated Hospital of Nanchang University.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of the Second Affiliated Hospital of Nanchang University, with a waiver of informed consent due to its retrospective design. All procedures involving human participants were conducted in accordance with the 1964 Declaration of Helsinki and its later amendments.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.