Dual-layer spectral detector computed tomography for adrenal adenoma characterization: radiation dose reduction and quantitative agreement of multiphase virtual noncontrast with true noncontrast imaging

Abstract

Background

Computed tomography (CT) is the preferred imaging modality for evaluating adrenal lesions; however, the associated radiation exposure remains a significant concern. Dual-layer spectral detector CT (SDCT)-derived virtual noncontrast (VNC) images may reduce radiation exposure by eliminating dedicated noncontrast scans, yet their agreement with true noncontrast (TNC) imaging remains debated. This study aimed to quantitatively evaluate the agreement and image quality of VNC images [reconstructed from the arterial phase (VNCa) and portal venous phase (VNCp)] compared to TNC images in adrenal adenomas stratified by lipid content, and to assess the radiation dose reduction.

Methods

A total of 103 patients with adrenal adenomas treated at the Adrenal Disease Center of West China Hospital of Sichuan University between March 2023 and September 2024 were enrolled in this prospective study. All patients underwent dual-layer SDCT examination, including TNC and arterial and venous phase scans. VNC images were reconstructed from contrast-enhanced phases. Objective metrics, including CT attenuation value [Hounsfield units (HU)], noise (standard deviation), signal-to-noise ratio (SNR), contrast-to-noise ratio, and absolute attenuation error, and subjective image quality were compared. Interobserver agreement was assessed through the calculation of interclass correlation coefficients. For objective and subjective comparisons between TNC and VNC images, statistical analyses were performed with paired t-tests and Wilcoxon signed-rank tests. The radiation dose with and without TNC was calculated.

Results

This study included 103 patients (48 males and 55 females) with a mean age of 51.33±12.55 years. A total of 123 adrenal adenomas were identified, including 28 lipid-rich adenomas and 95 lipid-poor adenomas. For lipid-poor adenomas, VNC and TNC images showed excellent agreement in CT attenuation values (P>0.05), and compared to VNCp images, VNCa images exhibited significantly lower noise (17.44±3.39 vs. 18.64±2.91 HU; P<0.001) and higher SNR (1.68±0.76 vs. 1.55±0.67; P<0.001). In lipid-rich adenomas, VNC images overestimated CT attenuation, showing high absolute attenuation errors (VNCaerror: 9.92±6.49 HU; VNCperror: 8.50±5.17 HU), although these remained within the acceptable threshold of ≤10 HU. In the subjective scores of image quality, TNC images outperformed VNC images [TNC: median 5, interquartile range (IQR) 5–5; VNC: median 5 (IQR 4–5); P<0.001], although VNC scores remained high. No significant statistical difference was observed between the VNCa and VNCp scores (P>0.05). For most of the surrounding nonadenoma tissues, VNC and TNC images demonstrated good agreement, with attenuation differences consistently within ≤10 HU. Replacing TNC images with VNCa images could reduce the effective dose by approximately 32.63% for lipid-poor adenomas.

Conclusions

Our findings suggest that for lipid-poor adenomas, VNCa demonstrates high agreement with TNC and provides superior image quality, supporting its use as a TNC substitute for reduced radiation dose. For lipid-rich adenomas, VNC should be applied with caution due to the potential risk of attenuation overestimation. Subtype classification remains essential in such studies.

Introduction

Adrenal adenomas are among the most common benign tumors of the adrenal gland, frequently detected incidentally during imaging studies performed for unrelated conditions. Although the majority of these tumors are benign and nonfunctional, accurate differentiation from other adrenal pathologies, such as pheochromocytomas, adrenocortical carcinomas, or metastatic lesions, is critical for guiding appropriate clinical management (1-3). Computed tomography (CT) is the imaging modality of choice for evaluating adrenal lesions due to its high spatial resolution and ability to provide detailed anatomical and functional information. Unenhanced CT [i.e., true noncontrast (TNC) CT], particularly through the measurement of Hounsfield units (HU), plays a pivotal role in characterizing adrenal adenomas. According to the European Society of Endocrinology Clinical Practice guidelines, homogeneous lesions with HU values ≤10 should generally be considered lipid-rich, benign adenomas, requiring no further imaging (1,4). However, for lesions with HU values >10, additional contrast-enhanced or multiphase CT scans are often necessary to exclude malignancy (1,5-7). Despite the diagnostic value of CT, the associated radiation exposure remains a significant concern, particularly for radiation-sensitive populations or patients requiring long-term follow-up for nonadrenal conditions. Dual-energy CT techniques have emerged as promising solutions, demonstrating improved adrenal lesion characterization while providing a reduction in radiation exposure (8-10).

In recent years, dual-layer spectral detector CT (SDCT) has emerged as an advanced imaging technology that offers the potential to reduce the radiation dose. By acquiring high- and low-energy data, reconstructing virtual noncontrast (VNC) images, and subtracting iodine attenuation from contrast-enhanced data, dual-layer SDCT can generate images that closely resemble traditional TNC scans, thereby eliminating the need for separate noncontrast scans and reducing the exposure of patients to radiation (11-16). Although the technical feasibility of VNC in abdominal imaging has been validated (12,15-20), its clinical utility for characterizing adrenal lesions remains controversial. Previous studies investigating the diagnostic performance of dual-energy CT-derived VNC for adrenal lesions have reported heterogeneous results (21-24). These discrepancies may stem from limitations such as small sample sizes, retrospective designs, or a narrow focus on specific adenoma subtypes (e.g., lipid-rich or lipid-poor adenomas), which fail to account for the full spectrum of adrenal lesion variation and phase-specific VNC reconstruction protocols. Notably, evidence indicates that VNC imaging tends to overestimate the CT attenuation values of fat (14,15). Given that adrenal adenomas are histologically classified into lipid-rich and lipid-poor subtypes based on their intracellular fat composition, the substantial variability in fat content between these subtypes raises critical questions regarding the reliability of VNC measurements. To address this issue, our study specifically conducted stratified analyses of these two distinct adenoma subtypes to evaluate potential disparities in VNC accuracy.

This prospective study evaluates VNC images derived from the arterial phase (VNCa) and portal venous phase (VNCp) across lipid-rich and lipid-poor subtypes through rigorous quantitative and qualitative comparisons with TNC. Our objectives were as follows: (I) to quantify the agreement between VNC and TNC measurements in different adenoma subtypes and (II) to identify optimal phase-specific VNC reconstruction protocols (arterial vs. portal venous) that optimize image quality and diagnostic consistency. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-854/rc).

Methods

Study population

This study was conducted at the Adrenal Disease Center of West China Hospital of Sichuan University, with 123 patients being enrolled between March 2023 and September 2024. Following screening based on inclusion and exclusion criteria, a total of 103 patients with adrenal adenoma were ultimately included. The inclusion criteria were (I) a previous diagnosis or high suspicion of adenomas and (II) age ≥18 years. There were distinct diagnostic criteria for lipid-rich and lipid-poor adrenal adenomas. (I) Lipid-rich adenomas were defined as lesions with CT attenuation values ≤10 HU on unenhanced CT. Meanwhile, (II) lipid-poor adrenal adenomas (CT attenuation >10 HU on unenhanced CT) were diagnosed if at least one of the following criteria were met: (i) demonstrated stability (no interval growth) over ≥6 months on prior or follow-up imaging; (ii) definitive imaging features consistent with adenoma on correlative studies (e.g., contrast-enhanced CT or magnetic resonance imaging); and (iii) histopathological confirmation (5,25,26). The exclusion criteria were as follows: (I) poor image quality due to respiratory motion artifacts, which precluded quantitative analysis; (II) the presence of other adrenal lesions, such as adrenal cortical carcinoma, pheochromocytoma, or other nonadenomatous pathologies; and (III) contraindications to contrast agents, including allergies to CT contrast agents or other medical contraindications. The screening process is illustrated in Figure 1.

Figure 1.

Flowchart of patient inclusion. CT, computed tomography; HU, Hounsfield units.

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by the Biomedical Ethics Committee of West China Hospital of Sichuan University (approval No. 2019 145). All patients who participated in the study provided written informed consent prior to undergoing the examination.

Dual-layer SDCT scanning protocol

All patients were scanned in the supine position with a dual-layer SDCT scanner (Spectral CT 7500; Philips Healthcare, Best, the Netherlands) for a triple-phase CT protocol, including TNC and arterial and venous phase images. The process began with an anteroposterior scout scan, followed by a TNC phase scan covering the adrenal glands and kidneys. Subsequently, a total-body-weight-adjusted dose of 1.2–1.5 mL/kg of intravenous contrast agent was administered at a flow rate of 3–5 mL/s with a high-pressure injector system. Bolus tracking, with a 150 HU threshold in the abdominal aorta, triggered the arterial scan at peak contrast concentration, while the venous phase started 70 seconds postinjection. The scan range remained consistent across all three phases. The scanning parameters were as follows: tube voltage, 120 kVp; tube current, modulated using automatic exposure control; collimation, 128 mm × 0.625 mm; field of view, 350 mm × 350 mm; matrix, 512×512; pitch, 1.000; rotation time, 0.5 seconds; and filter, standard (B). Details of the scanning protocol are summarized in Table 1.

Table 1. Detailed parameters of the scanning protocol.

Parameter	Value
Tube voltage	120 kVp
Tube current modulation	Automatic exposure control
Collimation	128 mm × 0.625 mm
Field of view	350 mm × 350 mm
Matrix size	512 × 512
Pitch	1
Rotation time	0.5 s
Reconstruction filter	Standard (B)
Contrast flow rate	3–5 mL/s
Total contrast volume	1.2–1.5 mL/kg (body weight-adjusted)
Bolus tracking threshold	150 HU (abdominal aorta)
Scan delay (arterial phase)	Triggered by bolus tracking
Scan delay (venous phase)	70 s after injection
Effective dose	k × DLP (k =0.015)

DLP, dose-length product; HU, Hounsfield units.

Image reconstruction

The acquired datasets were reconstructed into axial images with a slice thickness and interval of 1 mm. Conventional images were reconstructed with the iDose4 algorithm (Philips Healthcare), while spectral-based images were reconstructed with a spectral level 4 algorithm. VNC images were generated from contrast-enhanced datasets, with the arterial phase enhanced datasets used to create VNCa images and the venous phase datasets used to create VNCp images on a spectral post-processing workstation (IntelliSpace Portal; Philips Healthcare).

Objective evaluation of image quality

Two independent observers, an expert with 10 years of experience in abdominal imaging and a master’s student in medical imaging, performed the objective evaluation. CT attenuation and standard deviation (SD) were measured in several representative regions of interest (ROI) on TNC and VNC images. To validate the generalizability of the VNC algorithm across different abdominal tissues, the selected ROIs included adrenal adenomas, normal adrenal glands, subcutaneous fat, muscle, liver, spleen, pancreas, renal cortex, renal medulla, renal arteries, renal veins, and the abdominal aorta. The following steps were followed to ensure consistent and reliable measurements: (I) for adrenal adenomas, the largest cross-sectional area was selected, and circular ROIs covering two-thirds of the adenoma were placed, with areas of liquefaction, necrosis, calcification, or vessels being avoided. Subcutaneous fat, muscle, liver, spleen, and pancreas were measured at the same level as the adenoma, with circular ROIs of approximately 100 mm²; (II) for normal adrenal glands, the largest cross-sectional area of the contralateral adrenal gland was selected, and circular ROIs covering two-thirds of the short axis were placed; (III) at the renal hilum center slice, elliptical ROIs covering two-thirds of the renal cortex, medulla, and short axis for the renal vein were placed; (IV) at the level where the renal arteries originate from the abdominal aorta, circular ROIs (100 mm²) for the aorta and elliptical ROIs covering two-thirds of the short axis for the renal arteries were placed; (V) all measurements were obtained at the central slice and two adjacent slices above and below, with the average of the three slices used as the final value. For bilateral structures (e.g., renal cortex, medulla, arteries, and veins), the average of the left and right sides was used. Final values for each ROI were calculated from the average measurements obtained from the two observers. ROIs were initially placed on arterial or portal venous phase images and then copied to TNC and VNC images, with manual adjustments made to account for potential differences due to patient respiration and motion. The method of ROI placement is illustrated in Figure 2.

Figure 2.

Schematic representation of ROI placement. Green circles denote ROI. ROIs initially defined on arterial or portal venous phase images were automatically propagated to the corresponding TNC and VNC images. (A) Adrenal adenomas, muscle, subcutaneous fat, liver, spleen, and pancreas. (B) The abdominal aorta and right renal artery. (C) The left renal artery. (D) Normal adrenal glands. (E) The right renal vein, renal cortex, and medulla. (F) The left renal vein, renal cortex, and medulla. ROI, region of interest; TNC, true noncontrast; VNC, virtual noncontrast.

After obtaining the CT attenuation and SD values, we further evaluated image quality by calculating the absolute attenuation error (20), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR), consistent with the methodology employed in our previous work (27).

1. The absolute attenuation error for VNCa (VNCa_error) was calculated as the difference in CT attenuation values between the VNCa images and the TNC images, expressed as follows:

   $$VNC{a}_{error}=\left|H{U}_{VNCa}-H{U}_{TNC}\right|$$ [1]

   Where HU_VNCa is the CT attenuation of the VNCa images, and HU_TNC is the CT attenuation of the TNC images.

2. Similarly, the absolute attenuation error for VNCp (VNCp_error) between the VNCp images and the TNC images was calculated as follows:

   $$VNC{p}_{error}=\left|H{U}_{VNCp}-H{U}_{TNC}\right|$$ [2]

   An absolute attenuation error between the VNC and TNC images of 10 HU or less was considered acceptable (14,28).

3. The SNR was determined as follows:

   $$SNR=H{U}_{ROI}/S{D}_{ROI}$$ [3]

   Where HU_ROI is the CT attenuation of the ROI, and SD_ROI is the SD of the ROI.

4. The CNR was calculated as follows:

   $$CNR=(H{U}_{ROI}-H{U}_{muscle})/S{D}_{muscle}$$ [4]

   Where HU_ROI is the CT attenuation of the ROI, HU_muscle is the CT attenuation of muscle, and SD_muscle is the SD of muscle.

Subjective evaluation of image quality

Two independent readers, each with 15 years of experience in abdominal imaging, conducted the subjective assessment. Both reviewers were blinded to the specific reconstruction methods. Images were rated on a five-point scale based on anatomical detail clarity, artifact presence, and noise levels. The scoring scheme was as follows: 5= excellent, 4= good, 3= moderate, 2= poor, and 1= nondiagnostic.

Radiation dose assessment

The CT scanning protocol for all patients included both the TNC phase and the dual-phase contrast-enhanced scan. The dose-length product (DLP) (mGy∙cm) and effective dose (ED) (mSv) were recorded for each phase. The ED was the DLP registered by the CT scanner multiplied by a conversion coefficient (k) of 0.015 mSv/(mGy∙cm) (12), which can be expressed as follows: ED = k × DLP, where k is 0.015.

Statistical analysis

The Kolmogorov-Smirnov test was used to assess variable distributions, with normally distributed data reported as the mean ± SD and nonnormally distributed data as the median and interquartile range (IQR). Due to the nonrandom selection of evaluators, interobserver agreement was evaluated via the intraclass correlation coefficient (ICC) (type 3), with agreement levels classified as poor (0≤ ICC ≤0.2), fair (0.2< ICC ≤0.4), moderate (0.4< ICC ≤0.6), good (0.6< ICC ≤0.8), or excellent (0.8< ICC ≤1.0). Comparisons were conducted via paired t tests and Wilcoxon signed-rank tests, and Bland-Altman analysis was employed to evaluate the agreement in mean CT attenuation between TNC and VNC images. All statistical analyses were performed with SPSS 25.0 (IBM Corp., Armonk, NY, USA) and MedCalc version 20.218 (MedCalc Software, Ostend, Belgium), with statistical significance set at a two-sided P value <0.05.

Results

Patient characteristics

A total of 123 patients previously diagnosed with or suspected of adrenal adenomas were enrolled between March 2023 and September 2024. Of these patients, 20 were excluded due to poor image quality from respiratory motion artifacts (n=6), nonadenomatous adrenal lesions [adrenal cortical carcinoma (n=1), pheochromocytoma (n=4), or other pathologies (n=2)], or contrast agent contraindications (n=7), leaving 103 eligible cases (48 males and 55 females, with a mean age of 51.33±12.55 years). Among them, 20 had bilateral adenomas and 83 had unilateral adenomas. A total of 123 adrenal adenomas were identified, including 28 lipid-rich adenomas (CT attenuation ≤10 HU) and 95 lipid-poor adenomas (CT attenuation >10 HU).

Interobserver agreement

The interobserver agreement was consistently strong, ranging from good to excellent, as indicated by ICC values between 0.748 and 0.977. Detailed ICC results for various tissues and subjective scores are presented in Table S1.

Comparison of the mean CT attenuation (HU) values and imaging noise (SD) between TNC and VNC images

For lipid-rich adenomas, normal adrenal glands, renal arteries, and subcutaneous fat, VNC images demonstrated significantly higher mean CT attenuation values as compared to TNC images. In contrast, TNC images exhibited higher mean CT attenuation values in muscle, liver, spleen, pancreas, and renal medulla as compared to VNC images. No significant differences in mean CT attenuation values were observed between TNC and VNC images for lipid-poor adenomas or renal veins. Notably, there was no significant difference in mean CT attenuation values between TNC and VNCa images for the renal cortex and abdominal aorta, but TNC values were significantly higher than those of VNCp. A detailed summary of the comparisons is provided in Table 2. Figure 3 illustrates the phase-specific CT attenuation profiles of adrenal adenomas across the TNC, VNC, and contrast-enhanced phases.

Table 2. Comparison of the mean CT attenuation (HU) values between TNC and VNC images.

Region	TNC (HU)	VNCa (HU)	VNCp (HU)	P
				TNC vs. VNCa	TNC vs. VNCp	VNCa vs. VNCp
Lipid-rich adenomas	6.43±2.43	15.32±8.49	13.66±7.85	<0.001***	<0.001***	>0.05
Lipid-poor adenomas	28.58±11.34	28.12±10.96	27.72±9.78	>0.05	>0.05	>0.05
Normal adrenal glands	28.48±7.99	31.69±7.64	30.37±7.47	<0.001***	0.001**	0.01*
Subcutaneous fat	−110.73 (−115.37 to −105.17)	−107.70 (−111.37 to −101.73)	−105.60 (−109.17 to −100.57)	<0.001***	<0.001***	<0.001***
Muscle	54.47 (52.40–56.53)	48.49±4.86	49.50 (45.57–51.57)	<0.001***	<0.001***	>0.05
Liver	65.67 (60.77–70.77)	57.43 (53.40–63.33)	58.37 (54.67–62.80)	<0.001***	<0.001***	0.018*
Spleen	56.94±3.83	48.53 (46.30–50.13)	48.30 (46.13–50.77)	<0.001***	<0.001***	>0.05
Pancreas	48.37 (42.07–51.40)	41.93±5.76	41.27 (37.60–46.20)	<0.001***	<0.001***	>0.05
Renal cortex	35.31±3.36	35.58±4.87	34.05±4.19	>0.05	0.011*	0.011*
Renal medulla	37.15±4.51	27.50±4.56	31.14±5.05	<0.001***	<0.001***	<0.001***
Renal vein	43.89±4.38	44.89±5.09	42.90 (39.85–47.05)	>0.05	>0.05	>0.05
Renal artery	43.63±5.75	50.22±8.08	45.81±5.98	<0.001***	0.003**	<0.001***
Abdominal aorta	46.65±4.43	47.11±5.35	43.30±4.51	>0.05	<0.001***	<0.001***

Data are expressed as mean ± SD or median (interquartile range). Significance levels: *P<0.05, **P<0.01, ***P<0.001. CT, computed tomography; HU, Hounsfield units; SD, standard deviation; TNC, true noncontrast; VNC, virtual noncontrast; VNCa, virtual noncontrast image from the arterial phase; VNCp, virtual noncontrast image from the portal venous phase.

Figure 3.

Comparative analysis of CT attenuation in adrenal adenomas stratified by HU thresholds. Green circles mark ROIs for attenuation measurement. (A) Lipid-rich adenoma (CT attenuation ≤10 HU) and (B) lipid-poor adenoma (CT attenuation >10 HU). ROIs were consistently applied across all imaging phases to ensure comparative accuracy. Representative axial (a) true noncontrast, (b) arterial phase, and (c) portal venous phase images. (d) Virtual noncontrast image from the arterial phase and (e) virtual noncontrast image from the portal venous phase. Av, average; CT, computed tomography; HU, Hounsfield units; ROI, region of interest; SD, standard deviation.

For most tissues, VNC images exhibited lower SD values as compared to TNC images. However, for lipid-rich adenomas and normal adrenal glands, VNCa images demonstrated significantly lower SD values than did TNC images, whereas no statistically significant differences in SD values were observed between the VNCp and TNC images. Similarly, renal arteries showed no significant differences in SD values between TNC and VNC images. The details of the comparisons are shown in Table 3.

Table 3. Comparison of the imaging noise (HU) between TNC and VNC images.

Region	TNC (HU)	VNCa (HU)	VNCp (HU)	P
				TNC vs. VNCa	TNC vs. VNCp	VNCa vs. VNCp
Lipid-rich adenomas	19.69±4.75	17.47±3.59	18.47±3.21	0.003**	>0.05	>0.05
Lipid-poor adenomas	20.70±3.99	17.44±3.39	18.64±2.91	<0.001***	<0.001***	<0.001***
Normal adrenal glands	17.65±4.73	15.21±4.49	17.31±4.13	<0.001***	>0.05	<0.001***
Subcutaneous fat	11.80 (11.03–13.67)	10.13 (9.30–11.50)	10.50 (9.60–12.23)	<0.001***	<0.001***	>0.05
Muscle	17.13 (15.67–18.67)	13.95±2.49	14.39±2.34	<0.001***	<0.001***	0.016**
Liver	16.10±2.49	13.68±2.25	13.87 (12.50–15.50)	<0.001***	<0.001***	0.003**
Spleen	16.19±2.73	13.47±2.86	13.74±2.80	<0.001***	<0.001***	>0.05
Pancreas	19.73±3.09	17.79±2.91	18.00 (15.83–20.03)	<0.001***	<0.001***	0.09
Renal cortex	17.18±3.01	14.47±2.68	15.30 (13.85–17.25)	<0.001***	<0.001***	<0.001***
Renal medulla	17.05 (14.90–19.15)	14.70 (12.90–16.95)	16.04±2.80	<0.001***	<0.001***	0.005**
Renal vein	19.70±4.01	16.84±2.84	17.65±3.03	<0.001***	<0.001***	0.017*
Renal artery	17.59±4.09	17.38±3.74	17.74±3.65	>0.05	>0.05	>0.05
Abdominal aorta	21.93 (18.97–23.60)	19.24±2.78	20.03±2.45	<0.001***	<0.001***	0.002**

Data are expressed as mean ± SD or median (interquartile range). Imaging noise was quantified as the SD of CT attenuation values. Significance levels: *P<0.05, **P<0.01, ***P<0.001. CT, computed tomography; HU, Hounsfield units; SD, standard deviation; TNC, true noncontrast; VNC, virtual noncontrast; VNCa, virtual noncontrast image from the arterial phase; VNCp, virtual noncontrast image from the portal venous phase.

Comparison of the SNR and CNR between TNC and VNC images

For lipid-rich adenomas, lipid-poor adenomas, muscle, renal cortex, and renal veins, VNC images demonstrated significantly higher SNR values compared to TNC images. In contrast, subcutaneous fat and renal medulla exhibited higher SNR values on TNC images than on VNC images. No statistically significant differences in SNR were observed between TNC and VNC images for the liver and spleen. Notably, in normal adrenal glands, renal arteries, and the abdominal aorta, VNCa images showed higher SNR values than did TNC images, while no significant differences were found between VNCp and TNC images. The details of the comparisons are shown in Table S2.

For normal adrenal glands, the renal cortex, renal veins, renal arteries, and abdominal aorta, VNC images demonstrated higher CNR values as compared to TNC images. In contrast, TNC images exhibited higher CNR values than did VNC images in subcutaneous fat, pancreas, and the renal medulla. No statistically significant differences in CNR values were observed between the TNC and VNC images for adrenal adenomas. For the liver, VNCp images showed higher CNR values than did TNC, while no significant differences were found between the VNCa and TNC images. In the pancreas, TNC images demonstrated higher CNR values than did VNCa, whereas no significant differences were observed between the VNCp and TNC images. The details of the comparison results are shown in Table S3.

Comparison of absolute attenuation error between the VNCa and VNCp images

The mean absolute attenuation errors of the VNC images for all tissues were consistently ≤10 HU. Meanwhile, Bland-Altman plots (Figure 4) also indicated good agreement in CT attenuation between the TNC and VNC images across all ROI. For most tissues, there was no significant difference in mean absolute attenuation errors between the VNCa and VNCp images (P>0.05). However, the errors for fat were higher in the VNCp images, while the errors for the liver, renal medulla, and renal arteries were higher in the VNCa images. Further details are provided in Table 4.

Figure 4.

Bland-Altman plots demonstrating the agreement of CT attenuation measurements between TNC images and VNC images across all ROI. (A1,A2) Lipid-poor adrenal adenoma; (B1,B2) lipid-rich adrenal adenoma; (C1,C2) normal adrenal glands; (D1,D2) renal cortex; (E1,E2) renal medulla; (F1,F2) renal artery; (G1,G2) renal vein; (H1,H2) muscle; (I1,I2) subcutaneous fat; (J1,J2) abdominal aorta; (K1,K2) liver; (L1,L2) spleen; (M1,M2) pancreas. CT, computed tomography; HU, Hounsfield unit; ROI, region of interest; SD, standard deviation; TNC, true noncontrast; VNCa, virtual noncontrast image from the arterial phase; VNCp, virtual noncontrast image from the portal venous phase.

Table 4. Comparison of absolute attenuation error between VNCa and VNCp images.

Region	VNCaerror (HU)	VNCperror (HU)	P†
Lipid-rich adenomas	9.92±6.49	8.50±5.17	>0.05
Lipid-poor adenomas	5.57 (2.57–8.47)	5.10 (1.97–7.70)	>0.05
Normal adrenal glands	3.47 (1.63–6.57)	3.93 (2.00–7.30)	>0.05
Subcutaneous fat	4.71±2.85	5.88±3.67	<0.001***
Muscle	6.06±4.00	6.05±3.56	>0.05
Liver	7.72±4.22	7.09±4.31	0.018*
Spleen	8.23 (5.80–11.57)	8.84±4.67	>0.05
Pancreas	5.93 (3.87–9.07)	6.57 (3.70–9.03)	>0.05
Renal cortex	4.15 (2.00–6.35)	3.15 (1.50–5.85)	>0.05
Renal medulla	10.00±5.14	7.35 (3.45–10.05)	<0.001***
Renal vein	3.70 (1.95–6.60)	4.50 (2.00–7.15)	>0.05
Renal artery	7.20 (3.70–12.40)	5.30 (2.50–8.75)	<0.001***
Abdominal aorta	4.20 (1.97–6.90)	4.97 (2.17–9.10)	>0.05

Data are expressed as mean ± SD or median (IQR). ^†, comparison of VNCa_error and VNCp_error. Significance levels: *P<0.05, ***P<0.001. HU, Hounsfield unit; IQR, interquartile range; SD, standard deviation; VNCa, virtual noncontrast image from the arterial phase; VNCa_error, the absolute attenuation error between VNCa and TNC images; VNCp, virtual noncontrast image from the portal venous phase; VNCp_error, the absolute attenuation error between VNCp and TNC images.

Comparison of the SNR and CNR between the VNCa and VNCp images

For most tissues, the SNR values were higher in the VNCa images than in the VNCp images. However, for subcutaneous fat, the SNR values were higher in VNCp images. For the liver, spleen, renal medulla, and muscle, there was no significant difference in SNR values between the VNCa and VNCp images. The details of the comparisons are provided in Table S2.

For most tissues, there was no significant difference in CNR values between the VNCa and VNCp images. However, for subcutaneous fat and the renal medulla, the CNR values were higher in the VNCp images. For the renal arteries and abdominal aorta, the CNR values were higher in the VNCa images. The details of the comparisons are provided in Table S3.

Subjective image quality assessment by the two independent readers

Both readers consistently rated the TNC images higher, with a median score of 5 for both readers. The VNCa and VNCp images also received high scores, with a median of 5 (IQR 4–5) for both readers. However, there was a significant statistical difference in scores when TNC images (median 5, IQR 5–5) were compared to both the VNCa (median 5, IQR 4–5) and VNCp (median 5, IQR 4–5) images (P<0.001). There was also no significant difference in scores between the VNCa and VNCp images (P>0.05) for both readers. The results of the subjective evaluation of image quality are presented in Table 5.

Table 5. Subjective evaluation of image quality by two independent readers.

Score	TNC	VNCa	VNCp	P
				TNC vs. VNCa	TNC vs. VNCp	VNCa vs. VNCp
Reader 1				<0.001***	<0.001***	>0.05
Score
1	0	0	0
2	0	0	0
3	0	0	0
4	8	35	37
5	115	88	86
Median [IQR]	5 [5–5]	5 [4–5]	5 [4–5]
Reader 2				<0.001***	<0.001***	>0.05
Score
1	0	0	0
2	0	0	0
3	0	2	2
4	5	36	38
5	118	85	83
Median [IQR]	5 [5–5]	5 [4–5]	5 [4–5]

Significance levels: ***P<0.001. IQR, interquartile range; TNC, true noncontrast; VNCa, virtual noncontrast image from the arterial phase; VNCp, virtual noncontrast image from the portal venous phase.

Radiation dose evaluation

The ED values for the CT scan, including all phases, were 15.71±4.67 mSv. When the TNC phase was excluded from the CT scan, the ED value decreased to 10.57±3.13 mSv. This exclusion led to an approximate 32.63% reduction in the ED.

Discussion

This study aimed to evaluate the feasibility of VNC imaging in adrenal adenomas by conducting a quantitative and qualitative comparison of image quality between VNC images and traditional CT images. The results demonstrated that VNC imaging exhibits comparable performance to TNC imaging in evaluating lipid-poor adenomas. For lipid-poor adenomas, no statistically significant differences in CT attenuation were observed between the VNC and TNC images (P>0.05), confirming the reliability of VNC for these lesions. Notably, VNCa images exhibited superior image quality as compared to VNCp, with significantly lower noise (SD) and a higher SNR (P<0.001), supporting VNCa as the optimal choice as an alternative to TNC, as the effective radiation dose was reduced by 32.63% when dedicated TNC scans were omitted. For lipid-rich adenomas, VNC images showed significantly higher CT attenuation values than did TNC (P<0.001). While the mean absolute errors (VNCa: 9.92±6.49 HU; VNCp: 8.50±5.17 HU) were within the acceptable 10-HU threshold. This consistent overestimation suggests that the iodine subtraction algorithm may not be appropriate for low-density tissues. Caution is therefore required in the evaluation of lipid-rich adenomas on VNC images. For surrounding nonadenoma tissues, VNC and TNC demonstrated excellent agreement, with attenuation differences consistently ≤10 HU. Additionally, VNC images provided improved image quality metrics across most tissues, including reduced noise (SD) and enhanced SNR and CNR. These findings underscore the technical advantages of VNC in the comprehensive assessment of adrenal lesions and the characterization of surrounding tissue.

Previous studies have indicated a tendency for VNC images to overestimate attenuation values in adrenal lesions (21-24). This study further confirmed the presence of this phenomenon in lipid-rich adenomas but revealed divergent behavior in lipid-poor adenomas, consistent with a previous study (29) that reported no significant differences between VNC and TNC measurements. This discrepancy may be closely related to differences in intratumoral fat content. It has been established that VNC algorithms significantly overestimate CT attenuation values of fat (14,15,30), a finding corroborated by our observations: both the VNCa and VNCp images exhibited higher fat attenuation values as compared to TNC images. Given that lipid-rich adenomas are histologically defined by their elevated lipid content, the marked overestimation of CT attenuation values by VNC in these lesions likely stems from this inherent characteristic—specifically, the high lipid content of lipid-rich adenomas. In contrast, lipid-poor adenomas, which have lower lipid content, denser tissue structure, and stronger enhancement with higher iodine concentration during contrast-enhanced scans (Figure 3), demonstrate improved VNC–TNC consistency due to the precise removal of iodine signals. Although the diagnosis of lipid-rich adrenal adenomas relies on true unenhanced CT imaging, homogeneous lesions with attenuation ≤10 HU typically require no further imaging (1,4,5). However, our study revealed differences in agreement between the VNC and TNC CT attenuation values for lipid-rich and lipid-poor adenomas, highlighting the necessity of distinguishing these subgroups. Studies by Cao et al. (21) and Nagayama et al. (23), which did not differentiate adenoma subtypes and predominantly included lipid-rich cases (Cao et al.: 41 lipid-rich and 18 lipid-poor adenomas; Nagayama et al.: 57 lipid-rich and 47 lipid-poor adenomas), reported that VNC overestimates CT values for adenomas overall.

To determine the optimal VNC phase for TNC substitution, we conducted a comprehensive comparison of VNCa and VNCp images. For lipid-poor adenomas, compared to VNCp images, VNCa images demonstrated superior noise characteristics (lower SD and higher SNR) without significant differences in CT attenuation values, CNR, absolute attenuation error, or subjective image quality scores (P>0.05). These phase-dependent variations likely stem from differences in iodine concentration, hemodynamic characteristics, and the precision of iodine removal algorithms. During the arterial phase, the contrast agent has just entered the bloodstream, resulting in higher and more uniformly distributed iodine concentrations in blood vessels and tissues. This allows the iodine removal algorithm to more accurately and stably identify and remove iodine signals, thereby generating images with lower noise and a higher SNR. Typically, adenomas reach peak enhancement at 60 seconds, which is followed by rapid washout (3). In our study, portal venous phase images were acquired 70 seconds after injection, at which point the iodine concentration was relatively lower as compared to that in the arterial phase (Figure 3). This could have resulted in less precise and thorough removal of iodine signals in VNCp images, introducing more noise and reducing the SNR. However, the specific mechanisms and technical details of the iodine removal algorithm remain undisclosed, as they are proprietary intellectual property of the CT equipment manufacturers. Therefore, considering a balance of various parameters, we recommend using VNCa to replace TNC images for evaluating lipid-poor adenomas to effectively reduce the radiation dose. This is consistent with the findings reported in a colorectal cancer study (27) but conflicts with other studies that suggested the portal venous phase is a superior alternative to TNC (31) or that the differences between the arterial and portal venous phases are minimal (19). These discrepancies may arise from the following factors: (I) tissue-specific characteristics of iodine absorption and metabolism; (II) vendor-dependent algorithm heterogeneity, such as iodine removal strategies, postprocessing techniques, and noise suppression methods; (III) differences in scanner specifications, such as dual-energy technology types and detector performance; and (IV) variations in imaging protocols, including parameter settings and scanning conditions, which may affect the results.

Beyond lesion characterization, periadenoma tissue assessment is essential for differential diagnosis. We observed systematic directional variations in attenuation between VNC and TNC across different tissues, which suggests that a unified metric is needed for clinical applicability. To reconcile these discrepancies, we introduced the absolute attenuation error with a ≤10 HU acceptability threshold (14,20,28). Our analysis confirmed the clinically acceptable agreement for all tissues (mean error ≤10 HU), as indicated by Bland-Altman plots (Figure 4). Notably, renal medullary VNCa images showed the largest error (10.00±5.14 HU) due to the low perfusion and heterogeneous contrast distribution during the arterial phase, complicating iodine subtraction. Similarly, lipid-rich adenomas exhibited significant errors as intrinsic lipids interfered with the iodine differentiation algorithms. These findings highlight that the accuracy of VNC depends fundamentally on tissue-specific enhancement kinetics and noniodinated component composition.

Our study involved several limitations that should be addressed. First, the sample size was relatively small, particularly for the subgroup of lipid-rich adenomas (28 lipid-rich and 95 lipid-poor). Future studies with larger cohorts were needed to validate our conclusions. Second, while our inclusion criteria required prior confirmation or high suspicion of adenoma plus fulfillment of ≥1 criterion (imaging stability ≥6 months, definitive adenoma imaging features, or histopathological confirmation), the 6-month stability criterion specifically serves only to confirm benignity. It cannot definitively establish a diagnosis of adenoma subtype since benign adrenal lesions encompass multiple pathological entities. Future studies should integrate systematic histopathological correlation to validate imaging-based subtype classification. Third, the dual-layer SDCT system used in this study was from a single vendor, and further validation on other devices is required, as fundamental technical differences across dual-energy CT platforms may have implications for generalizability (15,32). Dual-layer SDCT systems acquire low- and high-energy data simultaneously through layered detectors, while dual-source systems use separate X-ray tubes operating at different kilovoltage peaks, which can introduce cross-scatter. Meanwhile, rapid tube voltage switching systems alternate kilovoltage settings during gantry rotation but may be subject to temporal misregistration. These variations in spectral separation mechanisms, temporal resolution, and material decomposition algorithms could influence the VNC reconstruction accuracy across platforms. Fourth, our study was limited to conventional-dose VNC reconstruction, and future studies should examine VNC performance in low-dose settings. Finally, we employed a single-center design, which may restrain the generalizability of the results, and validation in larger multicenter studies is required.

Conclusions

Our findings suggested that for lipid-poor adenomas, VNCa shows high agreement with TNC and superior image quality, supporting its use as a TNC substitute for reducing the radiation dose. VNC should be applied cautiously in lipid-rich adenomas due to the potential for attenuation overestimation, and subtype classification is critical in such studies.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by the Biomedical Ethics Committee of West China Hospital of Sichuan University (approval No. 2019 145). All patients who participated in the study provided written informed consent prior to undergoing examinations.

Data Sharing Statement

Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-854/dss

qims-15-09-7935-dss.pdf ^{(81.6KB, pdf)}

DOI: 10.21037/qims-2025-854