Abstract
Purpose
This study aimed to investigate the value of quantitative photon-counting computed tomography (PCCT) spectral parameters for the non-invasive assessment and prediction of programmed death-ligand 1 (PD-L1) expression in non-small cell lung cancer (NSCLC).
Methods
This single-center retrospective study enrolled 63 NSCLC patients who underwent pretherapeutic PCCT. Pretherapeutic PD-L1 expression was evaluated using immunohistochemistry and classified into three groups: negative (TPS < 1%), intermediate (1%–49%), and high (≥ 50%). The arterial and venous-phase CT values (A/V-40 keV, A/V-70 keV, and A/V-100 keV) and iodine concentration of the lesions were measured on 40, 70, and 100 keV monoenergetic images. Normalized iodine concentration and the slope of the spectral curve (λHU) were calculated accordingly. Spearman and Pearson correlation, receiver operating characteristic (ROC), univariate and multivariate logistic regression analyses were conducted to assess the correlations, predictive performance, and the independent predictive value of PCCT parameters for PD-L1 expression.
Results
Age and gender demonstrated statistically differences among the groups (P = 0.036; P = 0.008). A-40 keV and A-λHU also exhibited significant differences. Both parameters were negatively correlated with PD-L1 expression (A-λHU: rs = − 0.270, P = 0.032; A-40 keV: rs = − 0.279, P = 0.027). Multivariate logistic regression confirmed that age and A-λHU as independent predictors of intermediate and high PD-L1 expression, respectively (OR = 1.014 and 0.031; both P < 0.05). ROC analysis revealed that the clinical model, PCCT model, and combined model demonstrated favorable predictive performance for high PD-L1 expression, with area under the curve values of 0.762 (95%CI: 0.637–0.860), 0.846 (95% CI: 0.733–0.925), and 0.889 (95% CI: 0.785–0.954), respectively (all P < 0.001).
Conclusions
Age and A-λHU are valuable predictors of PD-L1 expression in NSCLC patients before treatment. The model, combining PCCT spectral parameters and clinical features, may serve as a non-invasive tool for individualized treatment in NSCLC.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s12880-026-02531-9.
Keywords: Photon-counting computed tomography, Spectral parameters, PD-L1, Non-small cell lung cancer
Introduction
Among all cancer types, lung cancer is one of the leading causes of cancer-related mortality globally, with non-small cell lung cancer (NSCLC) being the most common pathological subtype [1]. Despite recent advancements in diagnosis and therapy, many patients with NSCLC experience recurrence post-treatment, which portends a poor prognosis. Besides conventional modalities (e.g., surgery, radiotherapy, and chemotherapy), the application of immune checkpoint inhibitors has significant clinical benefits for patients [2]. With the development and application of immunotherapy, therapies targeting programmed death-ligand 1 (PD-L1) have significantly prolonged the overall survival of patients with NSCLC with manageable safety [3].
According to the most recent version of the National Comprehensive Cancer Network Clinical Practice Guidelines in Oncology, the PD-L1 expression status can be used to guide NSCLC treatment [4]. Currently, immunohistochemistry (IHC) is the gold standard for assessing PD-L1 expression status, but its invasive nature and poor reproducibility pose clinical challenges [5]. Imaging examinations, especially computed tomography (CT) and positron emission tomography/CT, enable noninvasive prediction of PD-L1 expression in patients with NSCLC and may reduce the frequency of unnecessary invasive biopsies [5–7]. Dual-energy CT overcomes the limitation that traditional CT provides only morphological information and allows the extraction of quantitative parameters via postprocessing to accurately assess iodine distribution and tumor angiogenesis status [8]. Studies have revealed that spectral CT imaging parameters can not only predict PD‑L1 expression but also assist in lung cancer diagnosis and the identification of metastatic lymph nodes [9–11].
Photon-counting CT (PCCT), a highly promising oncologic imaging modality, has opened a new avenue for tumor precision diagnosis and treatment [12]. Unlike energy-integrating detector CT, PCCT converts X-ray photons into electrical signals by generating electron–hole pairs and measuring individual photon energies, which eliminates the need for physical separation between detector elements, thereby achieving higher spatial resolution, lower image noise, and improved contrast and signal-to-noise ratio [13]. Meanwhile, PCCT enables spectral data acquisition and quantitative evaluation of tumors. Therefore, with its unique imaging advantages, PCCT compensates for the shortcomings of conventional imaging and enables a more accurate, noninvasive, and quantitative assessment of PD-L1 expression in patients with NSCLC.
PCCT presents significant advantages in tumor imaging; however, studies on the correlation between its spectral parameters and clinicopathological indicators, as well as immune marker expression in NSCLC, remain scarce, and its clinical application value warrants further exploration and verification. This study aimed to investigate the correlation between tumor PCCT spectral features and clinicopathological characteristics of NSCLC to provide evidence for the noninvasive prediction of PD-L1 expression in patients with NSCLC using PCCT, thereby facilitating earlier individualized and precise immunotherapy.
Materials and methods
Patients
This retrospective observational study obtained ethical approval from the Institutional Review Board of Shandong Cancer Hospital. All procedures involving human participants followed the ethical standards specified in the Declaration of Helsinki. This study was conducted at a single center.
We retrospectively reviewed the radiology database of 183 patients who were suspected of having pulmonary tumors and received pulmonary scanning with PCCT at our institution from May 2025 to February 2026. Among them, 63 patients (47 males and 16 females) were finally enrolled in this study. The inclusion criteria were as follows: (1) patients with NSCLC confirmed by surgical resection or biopsy; (2) patients who underwent PCCT examination within 2 weeks preoperatively or biopsy, with images of sufficient quality for diagnosis and analysis; (3) age of ≥ 18 years; (4) availability of complete clinical and pathological data. The exclusion criteria were as follows: (1) pathological diagnosis of SCLC or mixed SCLC; (2) prior preoperative antitumor treatments, including chemotherapy, radiotherapy, targeted therapy, or immunotherapy; (3) CT images with severe artifacts, motion artifacts, or excessive slice thickness, precluding tumor segmentation and feature extraction; (4) missing or uninterpretable pathological results for PD-L1 expression; and (5) comorbidities involving other malignant tumors. Finally, this study included 63 patients with NSCLC. Figure 1 illustrates the study flow chart and design.
Fig. 1.
Flowchart for patient enrollment
Data on age, gender, TNM stage, smoking history, serum tumor markers, and peripheral hematological parameters were collected from clinical medical records before any treatment, including carcinoembryonic antigen (CEA, reference range:0-5ng/ml), neuron-specific enolase (NSE, reference range:0-17ng/mL), cytokeratin 19 fragment (CYFRA21-1, reference range:0-3.3ng/mL), pro-gastrin-releasing peptide (ProGRP, reference range:0-65.7 pg/mL), squamous cell carcinoma–associated antigen (SCC-Ag, reference range:0-2.7 ng/mL), white blood cell count (WBC, reference range: 3.5–9.5 × 109/L), neutrophil count (Neu, reference range: 1.8–6.3 × 109/L), monocyte count (MO, reference range:0.10–0.60 × 109/L), absolute lymphocyte count (ALC, reference range:1.1–3.2 × 109/L), red blood cell count (RBC, reference range:3.8–5.1 × 1012/L), platelet count (PLT, reference range:125–350 × 109/L), hemoglobin (HGB, reference range:130–175 g/L), albumin (ALB, reference range:40–55 g/L), prealbumin (PAB, reference range:0.25–0.40 g/L), and hematocrit (Hct, reference range:40–50%). The neutrophil-to-lymphocyte ratio (NLR) was defined as Neu divided by ALC. The platelet-to-lymphocyte ratio (PLR) was calculated as PLT divided by ALC. The lymphocyte-to-monocyte ratio (LMR) was calculated as ALC divided by MO.
Complete blood counts were analyzed with a Sysmex XN9000 hematology analyzer. Biochemical parameters were determined on Hitachi 7600-020 and Roche Cobas 702 automatic biochemical analyzers. Tumor markers and immune indices (including CEA and NSE) were measured by electrochemiluminescence immunoassay (ECLIA) using Roche Cobas e601/e602 systems (Roche Diagnostics, Mannheim, Germany) with original reagents. All procedures complied with the manufacturer’s guidelines and laboratory standard operating procedures. The clinical laboratory is accredited to the ISO 15,189 standard for medical laboratories.
CT examination
All participants underwent contrast-enhanced chest CT using a clinical dual-source PCCT scanner (NAEOTOM Alpha; Siemens Healthineers, Forchheim, Germany). The examination was conducted at a tube voltage of 140 kVp and an image quality level of 80. Collimation was set as 144 × 0.4 mm, with a pitch factor of 0.8 and a gantry rotation time of 0.5s. Spectral post-processing (SPP) images and iodine maps were reconstructed with a 1-mm slice thickness, 0.7-mm increment, 3-grade quantum iterative reconstruction (QIR) level, and Qr40 kernel and transferred into the workstation (syngo. Via VB60; Siemens Healthineers, Forchheim, Germany) for further analysis.
A weight-adjusted dose of iodine-based contrast medium (370 mgI/mL; Bayer Healthcare, 1.2 mL/kg) was injected intravenously via a peripheral vein at a 2.5 mL/s flow rate. Bolus tracking was applied, with a region of interest (ROI) placed in the thoracic aorta at the heart level. Arterial phase images were scanned 25 s after the CT attenuation in the ROI that reached the threshold of 140 HU, and portal venous phase scanning was conducted 50 s afterward.
Imaging analysis
All data were processed and analyzed in the previously mentioned workstation. Two radiologists with 8 and 5 years of experience, respectively, independently delineated the ROI of the tumor, who were blinded to the patients’ clinical and pathological information, and reached a consensus on the image interpretations. Necrotic, cystic, and calcified areas were avoided during ROI1 delineation. Within the same imaging phase, ROI1 was placed on the largest axial slice of the lesion as well as on the two adjacent slices above and below (three consecutive slices in total). The mean value of these three slices was used as the final result. If the lesion was heterogeneous in density, the slice with the most solid component was selected for measurement. Furthermore, a circular ROI with an area of 5–10 mm2 (ROI2) was placed in the aorta or subclavian artery at the same level. Any discrepancies between the two radiologists were resolved by consensus through discussion.
Virtual monoenergetic images at 40, 70, and 100 keV in the arterial and venous phases were obtained, and CT values were measured in ROI1 (A-40, A-70, and A-100 keV and V-40, V-70, and V-100 keV). The slope of the spectral HU curve (λHU) was assessed using the equation λHU = (CT40 keV − CT100 keV) HU/(100 − 40) based on previous studies for both the arterial phase (A-λHU) and portal venous phase (V-λHU). Iodine concentrations of ROI2 in the arterial and venous phases (A-IC/V-IC) were measured on iodine maps. Additionally, the iodine concentration in the descending aorta or subclavian artery was measured in the same slice. The normalized iodine concentration (A-NIC and V-NIC), defined as the ratio of iodine concentration in the lesion to that in the descending aorta (NIC = IC/ICa), was calculated. These iodine concentration parameters (A-IC, V-IC, A-NIC, and V-NIC) were calculated in the arterial and venous phases, respectively. Effective atomic numbers (Z-effective [Zeff]) and dual-energy index (DEI) were measured with SPP images in the workstation for each ROI. PCCT-derived extracellular volume (ECV) was calculated using the formula: ECV = (1 − Hct) × (V-IC/V-ROI2) × 100%, where Hct refers to the volume fraction of red blood cells in the blood and V-ROI2 represents the arterial iodine concentration acquired during the vein phase of imaging.
Evaluation of PD-L1 expression by IHC
Tumor specimens were fixed in formalin, embedded in paraffin, and then sectioned into 3–5-µm-thick slices. The sections were stained with hematoxylin and eosin and observed under a light microscope. Additional sections were subjected to IHC staining using a PD-L1 22C3 antibody (Agilent-Dako). The tumor proportion score (TPS), which is defined as the percentage of tumor cells with partial or complete membranous staining relative to the total number of tumor cells in the sample, was used to quantify PD-L1 expression levels. Two pathologists, who were blinded to the patients’ clinical data, independently evaluated PD-L1 TPS levels. PD-L1 expression was classified into negative expression (PD-L1 TPS < 1%), intermediate expression (PD-L1 TPS = 1%–49%), and high expression (PD-L1 TPS ≥ 50%).
Statistical analysis
The Statistical Package for the Social Sciences software (Windows version 27.0; SPSS Inc., Chicago, IL, USA) was used for statistical analysis. GraphPad Prism 10 (GraphPad Software Inc., San Diego, CA, USA) was utilized to generate scatter plots. The one‑sample Shapiro–Wilk test was used to evaluate the normality of continuous data. Normally distributed data were reported as the mean ± standard deviation, and one‑way analysis of variance was used to compare the three groups, with Tukey’s honestly significant difference (HSD) method employed for post hoc multiple comparisons. The Dunn–Bonferroni method is used for post hoc pairwise comparisons. Non‑normally distributed data were expressed as the median (interquartile range), with the Kruskal–Wallis H test used for comparisons among the three groups, and Tukey’s HSD method employed for post hoc multiple comparisons. Categorical data were expressed as the number of cases, and the chi‑square test was applied for comparison. Receiver operating characteristic (ROC) curves were plotted to investigate the predictive efficacy of each quantitative parameter with statistically significant differences. Combined ROC analysis and the area under the curve (AUC) were applied to assess the predictive performance of the combined parameters. The Pearson product-moment correlation coefficient (r) or Spearman rank correlation coefficient (rs) was utilized to assess the correlations between tumor PD-L1 expression level and quantitative parameters of PCCT. Multinomial logistic regression analysis was conducted to ascertain the associations between gender, PCCT quantitative parameters, and PD-L1 expression. To prevent multicollinearity, collinearity diagnostics were performed through linear regression to calculate VIF and tolerance. VIF of > 10 indicated severe multicollinearity, and relevant variables were excluded for model stability. We adopted the Model Fitting Information, Pearson test, deviance test and Nagelkerke R² to evaluate the goodness-of-fit of multinomial logistic regression models, with a P-value > 0.05 indicating adequate model fit. All tests were two-sided, and a P-value of < 0.05 indicated statistical significance.
Results
Patient’s baseline information
The 63 enrolled patients had a mean age of 65.48 ± 8.27 years. Of them, 28, 22, and 13 had negative, intermediate, and high expressions, respectively (Fig. 2). Except for gender (P = 0.008) and age (P = 0.036), no significant differences in pathological type, clinical stage, smoking history, and laboratory examinations were found among groups with different PD‑L1 expression levels (all P > 0.05). Table 1; Fig. 3 show the clinical and pathological characteristics of patients with NSCLC with different PD-L1 expression levels.
Fig. 2.
PCCT parameters and pathological images of NSCLC patients with different PD-L1 expression levels. (A–D) Patient 1 with adenocarcinoma in the right middle lobe (high PD-L1 expression, TPS = 80%): 40 keV virtual monoenergetic, fusion, Zeff, and pathological images. (E–H) Patient 2 with adenocarcinoma in the right upper lobe (intermediate PD-L1 expression, TPS = 15%): 40 keV virtual monoenergetic, fusion, Zeff, and pathological images. (I–L) Patient 3 with squamous cell carcinoma in the left upper lobe (negative PD-L1 expression, TPS < 1%): 40 keV virtual monoenergetic, fusion, Zeff, and pathological images
Table 1.
Comparison of clinicopathological characteristics of NSCLC patients with different PD-L1 expression levels
| Characteristics | PD-L1 TPS < 1% | PD-L1 TPS 1–49% | PD-L1 TPS ≥ 50% | H/F/χ2 | P |
|---|---|---|---|---|---|
| Age | 62.32 ± 8.22 | 67.63 ± 8.05 | 66.25 ± 7.19 | 1.276 | 0.036 |
| Gender | 0.008 | ||||
| Male | 16 | 18 | 13 | ||
| Female | 12 | 4 | 0 | ||
| Histologic type | 0.325 | ||||
| Squamous cell carcinoma | 7 | 11 | 4 | ||
| Adenocarcinoma | 20 | 10 | 8 | ||
| Others | 1 | 1 | 1 | ||
| Stage | 0.143 | ||||
| I-II | 10 | 4 | 1 | ||
| III-IV | 18 | 18 | 12 | ||
| Smoking history | 0.831 | ||||
| Current smoker | 10 | 12 | 6 | ||
| Ex smoker | 5 | 2 | 1 | ||
| Never smoker | 12 | 7 | 5 | ||
| Unknown | 1 | 1 | 1 | ||
| Pretreatment laboratory parameters | |||||
| CEA (ng/mL) | 6.36 (3.23, 146.00) | 6.17 (3.35, 31.58) | 2.41 (1.59, 6.65) | 3.887 | 0.143 |
| NSE (ng/mL) | 17.90(14.30, 21.20) | 17.70 (13.35,20.57) | 22.10 (14.18, 27.45) | 1.699 | 0.428 |
| CYFRA21-1(ng/mL) | 6.38 (4.03, 16.30) | 6.82 (2.48, 12.57) | 6.97 (3.85, 27.75) | 1.665 | 0.435 |
| ProGRP(pg/mL) | 46.10 (40.40, 64.40) | 39.90 (33.15, 50.53) | 46.55 (36.83, 62.23) | 2.878 | 0.237 |
| SCC-Ag(ng/mL) | 1.19 (0.82, 1.90) | 1.34 (0.62, 5.22) | 1.08 (0.73, 1.43) | 0.806 | 0.668 |
| WBC (×109/L) | 6.48 (5.17, 9.67) | 6.94 (6.01, 9.64) | 6.28 (5.92, 8.83) | 0.949 | 0.622 |
| ALC(×109/L) | 1.42 (1.12, 2.12) | 1.55 (1.02, 2.14) | 1.68 (1.32, 2.21) | 0.433 | 0.805 |
| Neu(×109/L) | 5.30 ± 2.41 | 5.13 ± 1.44 | 4.64 ± 1.31 | 0.287 | 0.752 |
| MO(×109/L) | 0.48 ± 0.23 | 0.55 ± 0.20 | 0.62 ± 0.16 | 1.772 | 0.179 |
| RBC (×1012/L) | 4.60 ± 0.46 | 4.60 ± 0.66 | 4.46 ± 0.41 | 0.467 | 0.629 |
| HGB (×109/L) | 138.21 ± 14.60 | 138.13 ± 19.82 | 132.25 ± 15.56 | 0.767 | 0.469 |
| ALB (g/L) | 41.85 (38.18, 42.88) | 41.50 (38.55, 43.95) | 41.50 (38.95, 42.35) | 0.165 | 0.921 |
| PAB (g/L) | 0.23 ± 0.06 | 0.22 ± 0.06 | 0.23 ± 0.06 | 0.327 | 0.722 |
| Hct (%) | 41.84 ± 4.38 | 42.04 ± 5.90 | 40.44 ± 3.99 | 0.786 | 0.460 |
| PLT(×109/L) | 265.0 (220.0, 343.0) | 282.0 (200.75, 310.25) | 282.50 (249.25, 323.50) | 0.120 | 0.942 |
| NLR | 3.10 (2.34, 4.62) | 2.91 (2.51, 4.34) | 2.57 (2.08, 3.52) | 1.581 | 0.454 |
| PLR | 207.03 (150.67, 237.66) | 151.18 (131.99, 266.91) | 156.48 (124.49, 211.18) | 0.650 | 0.722 |
| LMR | 3.57 (2.87, 5.67) | 3.11 (2.14, 4.07) | 2.72 (2.51, 3.47) | 3.357 | 0.187 |
CEA, carcinoembryonic antigen; NSE, neuron-specific enolase; CYFRA21-1, cytokeratin 19 fragment; ProGRP, pro-gastrin-releasing peptide; SCC-Ag, squamous cell carcinoma–associated antigen; WBC, white blood cell count; Neu, neutrophil count; MO, monocyte count; ALC, absolute lymphocyte count; RBC, red blood cell count; PLT, platelet count; HGB, hemoglobin; ALB, albumin; PAB, prealbumin; Hct, hematocrit; NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; LMR, lymphocyte-to-monocyte ratio. Neu, MO, RBC, HGB, PAB and Hct were normally distributed, while WBC, ALC, ALB, PLT, NLR, PLR, LMR, CEA, NSE, CYFRA21-1, ProGRP and SCC-Ag exhibited non-normal distribution
Note: Clinicopathological characteristics among different PD-L1 expression groups were compared using Chi‑square test, Kruskal-Wallis H test or one-way analysis of variance (ANOVA). *P < 0.05
Fig. 3.
Differences in clinical characteristics and PCCT spectral parameters among PD-L1 TPS groups (< 1%, 1%–49%, and ≥ 50%). (A) Age, (B) A-λHU, (C) A-40 keV. *P < 0.05; ns: not significant
Comparison of quantitative spectral parameters
Except for A-40 keV (P = 0.017) and A-λHU (P = 0.037), no significant differences were observed in the remaining PCCT spectral parameters (Table 2; Fig. 3).
Table 2.
Comparison of quantitative PCCT spectral parameters in NSCLC patients with different PD-L1 expression levels
| Characteristics | PD-L1 TPS < 1% | PD-L1 TPS 1–49% | PD-L1 TPS ≥ 50% | H/F | P |
|---|---|---|---|---|---|
| A-IC | 0.73 ± 0.39 | 0.67 ± 0.26 | 0.55 ± 0.24 | 1.438 | 0.246 |
| A-NIC | 0.09 (0.06, 0.13) | 0.10 (0.07, 0.15) | 0.08 (0.05, 0.16) | 0.371 | 0.831 |
| A-40 keV | 88.50 (75.54, 119.04) | 91.62 (83.55, 105.30) | 76.25 (74.23, 80.31) | 8.128 | 0.017 |
| A-70 keV | 52.94 (48.24, 65.19) | 53.78 (49.17, 61.13) | 53.44 (48.83, 57.08) | 0.091 | 0.956 |
| A-100 keV | 45.62 ± 6.70 | 44.51 ± 6.88 | 46.24 ± 3.98 | 0.349 | 0.707 |
| A-λHU | 0.76 (0.50, 1.30) | 0.85(0.61, 0.96) | 0.57 (0.48, 0.68) | 6.593 | 0.037 |
| A-Zeff | 7.87 ± 0.24 | 7.83 ± 0.21 | 7.70 ± 0.13 | 0.289 | 0.065 |
| A-DEI | 0.0069 ± 0.0043 | 0.0064 ± 0.0034 | 0.0042 ± 0.0021 | 2.319 | 0.107 |
| V-IC | 1.25 ± 0.46 | 1.19 ± 0.40 | 1.09 ± 0.38 | 0.672 | 0.514 |
| V-NIC | 0.31 (0.25, 0.46) | 0.33 (0.22, 0.41) | 0.32 (0.21, 0.37) | 0.183 | 0.913 |
| V-40 keV | 141.21 ± 41.59 | 138.15 ± 36.64 | 126.36 ± 31.61 | 0.689 | 0.506 |
| V-70 keV | 68.92 ± 13.41 | 68.35 ± 12.47 | 66.77 ± 12.77 | 0.123 | 0.885 |
| V-100 keV | 50.52 ± 7.04 | 50.80 ± 7.86 | 51.96 ± 8.52 | 0.161 | 0.851 |
| V-λHU | 1.56 ± 0.64 | 1.46 ± 1.36 | 1.25 ± 0.42 | 1.342 | 0.269 |
| V-Zeff | 8.15 ± 0.29 | 8.11 ± 0.29 | 8.00 ± 0.16 | 1.318 | 0.275 |
| V-DEI | 0.0110 (0.0083, 0.0160) | 0.0110 (0.0078, 0.0135) | 0.0100 (0.0080, 0.0115) | 2.410 | 0.300 |
| ECV | 0.1766 (0.1531, 0.2549) | 0.1815 (0.1303, 0.2324) | 0.1953 (0.1361, 0.2184) | 0.048 | 0.976 |
A=arterial phase; ECV=extracellular volume; DEI=dual-energy index; IC=iodine concentrations; NIC= normalized iodine concentration; TPS= tumor proportion score; V=portal venous phase; Zeff= effective atomic numbers; λHU = the slope of the spectral HU curve. Note: PCCT spectral parameters among PD-L1 negative, intermediate, and high expression groups were compared using Kruskal-Wallis H test or one-way ANOVA. *P < 0.05
Further inter-group comparison revealed that A-40 keV was significantly lower in the PD-L1 high-expression group than in the negative-expression (P = 0.012) and intermediate-expression groups (P = 0.008), with no significant difference between the negative- and intermediate-expression groups (P = 0.773). Similarly, A-λHU was significantly lower in the PD-L1 high-expression group than in the negative- (P = 0.020) and intermediate-expression groups (P = 0.019), with no significant difference between the negative- and intermediate-expression groups (P = 1.000).
Correlation of quantitative spectral parameters with PD-L1 expression
Figure 4 shows the correlation between quantitative PCCT spectral parameters and different PD‑L1 expression groups. Weak negative correlations were observed between PD-L1 TPS expression and A-λHU (rs = − 0.270, P = 0.032), A-40 keV (rs = − 0.279, P = 0.027), A-Zeff (r = − 0.289, P = 0.022), and A-DEI (r = − 0.265, P = 0.036). The remaining PCCT spectral parameters demonstrated no significant correlations.
Fig. 4.
Correlation between PCCT spectral parameters and PD-L1 expression in NSCLC patients. (A–B) Spearman correlation coefficients (rs); (C–D) Pearson correlation coefficients
Predictive value of spectral parameters for PD-L1 expression
Univariate multinomial logistic regression identified five factors that are significantly associated with three different PD-L1 expression groups (all P < 0.05): age (χ² = 7.017, P = 0.029), gender (χ² = 12.293, P = 0.002), A-40 keV (χ² = 9.280, P = 0.010), A-λHU (χ² = 9.262, P = 0.009), and A-Zeff (χ² = 6.374, P = 0.041). The PD‑L1 high‑expression group included no female patients; thus, complete separation occurred in the multinomial logistic regression, which reduces model stability. Therefore, we conducted a sensitivity analysis by refitting the model without gender to reliably assess the independent predictive value of PCCT spectral parameters for PD‑L1 expression. Accordingly, gender was excluded from the multivariate logistic regression model to avoid overfitting.
Multicollinearity analysis between A-40 keV and A-λHU revealed that both parameters had VIF values > 10, indicating severe multicollinearity (Table S1). Age was included separately with A-40 keV, A-Zeff, and A-λHU in multivariate regression models. Using PD-L1 negative expression as the reference category and comparing it with the medium- and high-expression groups, the results indicated that A-40 keV (P = 0.016, odds ratio [OR] = 0.948, 95% confidence interval [CI]: 0.907–0.990), A-Zeff (P = 0.031, OR = 0.008, 95% CI: 0.000–0.639), and A-λHU (P = 0.013, OR = 0.031, 95% CI: 0.002–0.486) were all independent predictors of high PD-L1 expression (Table S2). Comprehensive comparison showed that A‑λHU had the strongest effect size and the smallest P value. Therefore, given its potential clinical value, we selected A‑λHU as the representative spectral parameter for future analysis.
Multivariate regression analysis revealed that, with PD‑L1 negative expression as the reference category, age was considered an independent predictor of intermediate PD‑L1 expression (TPS 1%–49% vs. <1%; OR = 1.104, 95% CI: 1.018–1.197, P = 0.017), whereas A‑λHU was independently associated with high PD‑L1 expression (TPS ≥ 50% vs. <1%; OR = 0.031, 95% CI: 0.002–0.486, P = 0.013). Table 3 summarizes the results.
Table 3.
Univariate and multinomial logistic regression analyses for predicting different PD-L1 expression in NSCLC
| Factor | PD-L1 expression comparison | Univariate Analysis Likelihood Ratio | Multivariate Analysis | ||
|---|---|---|---|---|---|
| χ² | P | OR (95% CI) | P | ||
| Age | TPS 1–49% vs. TPS < 1% | 7.017 | 0.029 | 1.104(1.018–1.197) | 0.017 |
| TPS ≥ 50% vs. TPS < 1% | 1.011(0.925–1.106) | 0.806 | |||
| Gender | TPS 1–49% vs. TPS < 1% | 12.293 | 0.002 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| A-IC | TPS 1–49% vs. TPS < 1% | 3.070 | 0.216 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| A-NIC | TPS 1–49% vs. TPS < 1% | 3.025 | 0.220 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| A-40 keV | TPS 1–49% vs. TPS < 1% | 9.280 | 0.010 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| A-70 keV | TPS 1–49% vs. TPS < 1% | 1.194 | 0.551 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| A-100 keV | TPS 1–49% vs. TPS < 1% | 0.730 | 0.694 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| A-λHU | TPS 1–49% vs. TPS < 1% | 9.262 | 0.010 | 0.638(0.143–2.855) | 0.557 |
| TPS ≥ 50% vs. TPS < 1% | 0.031(0.002–0.486) | 0.013 | |||
| A-Zeff | TPS 1–49% vs. TPS < 1% | 6.374 | 0.041 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| A-DEI | TPS 1–49% vs. TPS < 1% | 5.342 | 0.069 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| V-IC | TPS 1–49% vs. TPS < 1% | 1.421 | 0.491 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| V-NIC | TPS 1–49% vs. TPS < 1% | 0.060 | 0.970 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| V-40 keV | TPS 1–49% vs. TPS < 1% | 1.464 | 0.481 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| V-70 keV | TPS 1–49% vs. TPS < 1% | 0.259 | 0.879 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| V-100 keV | TPS 1–49% vs. TPS < 1% | 0.332 | 0.847 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| V-λHU | TPS 1–49% vs. TPS < 1% | 2.878 | 0.237 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| V-Zeff | TPS 1–49% vs. TPS < 1% | 2.794 | 0.247 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| V-DEI | TPS 1–49% vs. TPS < 1% | 4.071 | 0.131 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
| ECV | TPS 1–49% vs. TPS < 1% | 0.415 | 0.813 | Not entered | |
| TPS ≥ 50% vs. TPS < 1% | |||||
A=arterial phase; ECV=extracellular volume; DEI=dual-energy index; IC=iodine concentrations; NIC= normalized iodine concentration; TPS= tumor proportion score; V=portal venous phase; Zeff= effective atomic numbers; λHU = the slope of the spectral HU curve
Note: *P < 0.05, the p-value is derived from Univariate and multivariate logistic regression analysis
Further multivariate regression analysis with intermediate PD-L1 expression as the reference category (Table S3) revealed that both A-40 keV (OR = 0.955, 95% CI: 0.916–0.997, P = 0.039) and A-λHU (OR = 0.060, 95% CI: 0.004–0.871, P = 0.034) were independent predictors of high PD-L1 expression, among which A-λHU demonstrated superior statistical significance and a larger effect size.
Predictive model of quantitative spectral parameters for PD-L1 expression
A-40 keV and A-λHU demonstrated no significant differences between the negative and intermediate PD-L1 expression groups; thus, we reclassified PDL1 expression into two groups: low expression (< 1% and 1%–50%) and high expression (≥ 50%). ROC curve analysis was conducted to assess the ability of PCCT parameters for predicting PD-L1 expression in patients with NSCLC.
A-40 keV, A-λHU, A-Zeff, and A-DEI were predictive of PD-L1 expression, with AUCs of 0.757 (95% CI: 0.632–0.856, P < 0.001), 0.732 (95% CI: 0.606–0.836, P < 0.001), 0.704 (95% CI: 0.575–0.812, P = 0.0036), and 0.680 (95% CI: 0.550–0.792, P = 0.0139), respectively. The optimal cutoff values were ≤ 80.55 for A-40 keV (sensitivity, 84.6%; specificity, 74.0%), ≤ 0.78 for A-λHU (sensitivity, 100.0%; specificity, 67.0%), ≤ 7.7 for A-Zeff (sensitivity, 69.2%; specificity, 64.0%), and ≤ 0.004 for A-DEI (sensitivity, 61.5%; specificity, 70.0%). All other spectral parameters demonstrated no predictive value (all P > 0.05). Table 4; Fig. 5 show the results.
Table 4.
ROC curve analysis of PCCT spectral parameters for predicting high PD-L1 expression in NSCLC
| Parameters | AUC | Threshold | P | 95%CI | Sensitivity | Specificity | |
|---|---|---|---|---|---|---|---|
| Lower bound | Lower bound | ||||||
| A-40 keV | 0.757 | ≤ 80.55 | <0.001 | 0.632 | 0.856 | 84.62 | 74.00 |
| A-70 keV | 0.518 | ≤ 62.15 | 0.8190 | 0.389 | 0.646 | 100.00 | 24.00 |
| A-100 keV | 0.570 | > 41.31 | 0.3761 | 0.439 | 0.694 | 100.00 | 26.00 |
| A-λHU | 0.732 | ≤ 0.78 | <0.001 | 0.606 | 0.836 | 100.00 | 67.00 |
| A-IC | 0.641 | ≤ 0.6 | 0.0780 | 0.510 | 0.758 | 76.92 | 58.00 |
| A-NIC | 0.542 | ≤ 0.08 | 0.6876 | 0.411 | 0.668 | 61.54 | 56.00 |
| A-Zeff | 0.704 | ≤ 7.7 | 0.0036 | 0.575 | 0.812 | 69.23 | 64.00 |
| A-DEI | 0.680 | ≤ 0.004 | 0.0139 | 0.550 | 0.792 | 61.54 | 70.00 |
| V-40 keV | 0.606 | ≤ 136.92 | 0.2122 | 0.475 | 0.727 | 84.62 | 48.00 |
| V-70 keV | 0.545 | ≤ 68.33 | 0.6197 | 0.414 | 0.671 | 69.23 | 50.00 |
| V-100 keV | 0.537 | > 48.12 | 0.6948 | 0.407 | 0.664 | 69.23 | 44.00 |
| V-λHU | 0.663 | ≤ 1.33 | 0.0512 | 0.533 | 0.777 | 76.92 | 64.00 |
| V-IC | 0.606 | ≤ 1.2 | 0.3148 | 0.454 | 0.730 | 81.98 | 45.24 |
| V-NIC | 0.525 | ≤ 0.37 | 0.7821 | 0.396 | 0.653 | 84.62 | 38.00 |
| V-Zeff | 0.636 | ≤ 0.8 | 0.0628 | 0.505 | 0.754 | 69.23 | 58.00 |
| V-DEI | 0.625 | ≤ 0.014 | 0.0926 | 0.494 | 0.744 | 100.00 | 30.00 |
| ECV | 0.503 | ≤ 0.223 | 0.9734 | 0.374 | 0.632 | 84.62 | 38.00 |
| Clinical model | 0.762 | > 0.164 | <0.001 | 0.637 | 0.860 | 100.00 | 44.00 |
| PCCT model | 0.846 | > 0.166 | <0.001 | 0.733 | 0.925 | 92.31 | 70.00 |
| Combined model | 0.889 | > 0.122 | <0.001 | 0.785 | 0.954 | 100.00 | 70.00 |
A=arterial phase; ECV=extracellular volume; DEI=dual-energy index; IC=iodine concentrations; NIC= normalized iodine concentration; TPS= tumor proportion score; V=portal venous phase; Zeff= effective atomic numbers; λHU = the slope of the spectral HU curve; ROC, receiver operating characteristic; AUC. the area under curve; 95% CI, 95% confidence interval
Note: *P < 0.05, the P-value is derived from the receiver operating characteristic (ROC) curve analysis
Fig. 5.
ROC curves of PCCT spectral parameters (A) and three predictive models (B) for predicting PD-L1 high expression in patients with NSCLC
ROC curve analysis was conducted to assess the predictive ability of clinical characteristics (age and sex), PCCT spectral parameters (A-40 keV and A-λHU), and their combination for high PD-L1 expression, achieving AUCs of 0.762 (95% CI: 0.637–0.860, P < 0.001), 0.846 (95% CI: 0.733–0.925, P < 0.001), and 0.889 (95% CI: 0.785–0.954, P < 0.001). DeLong’s test revealed that the combined model was significantly superior to the clinical model (Z = 2.205, P = 0.027), demonstrating the best predictive performance. Figure 5 illustrates the results. To avoid the risk of overfitting due to the limited sample size (total n = 63, with only 13 patients in the PD‑L1 high‑expression subgroup), we performed internal validation of the prediction model using 1,000 bootstrap resampling. The optimism‑corrected AUC was 0.847 (95% bootstrap CI: 0.778–0.949).
Discussion
In our study, we investigated the value of quantitative PCCT spectral parameters for the non-invasive assessment and prediction of PD-L1 expression in NSCLC, which is essential for identifying patients who may benefit from PD-L1-targeted immunotherapy. No significant differences in pathological type, clinical stage, smoking history, or laboratory examinations were observed among the three PD-L1 expression groups (all P > 0.05), indicating that these factors may reliably predict PD-L1 expression in NSCLC, which is consistent with the results of the most recent retrospective studies [14, 15]. In contrast, age and gender demonstrated statistically significant differences in distribution among the three PD-L1 expression subgroups (P = 0.036 and 0.008, respectively). Further univariate regression analysis confirmed that age and gender were influencing factors for PD-L1 expression (P = 0.029 and 0.002, respectively). The PD-L1 high-expression group included no female patients; thus, complete separation occurred in the multinomial logistic regression, reducing model stability. Further multivariate regression analysis revealed that, including only gender and A-λHU, age was identified as an independent predictor of negative and intermediate PD-L1 expression in tumors (OR = 1.014, P = 0.017). These results indicate that age and gender are factors influencing tumor PD-L1 expression, a conclusion partially supported by previous studies [16–19].
The present study revealed that A-40 keV and A-λHU were the core spectral parameters for differentiating PD-L1 expression levels (P < 0.01). Further study demonstrated that A-40 keV and A-λHU were significantly lower in the high PD-L1 expression group than in the negative and intermediate expression groups, with no significant differences between the negative and intermediate expression groups. These findings indicate that A-40 keV and A-λHU are potential imaging markers for distinguishing high PD-L1 expression from negative and intermediate expression, but not for differentiating negative from intermediate expression. Correlation analysis revealed that A-λHU and A-40 keV were weakly negatively correlated with PD-L1 expression (rs = − 0.270, P = 0.032; rs = − 0.279, P = 0.027, respectively). Considering the severe multicollinearity among A-40 keV and A-λHU and comparisons after multivariate regression using different parameters (Table S1), the multivariate regression analysis finally included age and A-λHU. This approach ensured statistical rigor and eliminated confounding effects caused by collinearity. Multivariate logistic regression analysis confirmed A-λHU as an independent predictor of high PD-L1 expression (OR = 0.031, P = 0.013). Using the intermediate PD-L1 expression group as the reference, further multivariate logistic regression analysis showed that both A-40 keV and A-λHU enabled differentiation between the high- and intermediate-expression groups, with A-λHU demonstrating a substantially larger effect size. These results demonstrated that A-λHU and A-40 keV reflect iodine concentration and distribution within tumor tissues, as well as tumor microvascular density and permeability [10, 20, 21].
The following are possible mechanisms for the abovementioned results. High PD-L1 expression tumors typically demonstrate a more aggressive phenotype, where excessive proliferation of tumor cells leads to tissue hypoxia, thereby stimulating high expression of vascular endothelial growth factor and promoting tumor angiogenesis [22]. However, such tumor neovasculature frequently presents significant structural abnormalities due to the dysfunction of pericyte-like cells, leading to the widening of endothelial cell gaps and manifesting as increased vascular permeability [23, 24]. Alternatively, they may present as dysfunctional capillaries: blood within the tumor may be rapidly shunted through functional arteriovenous shunts, or their lumens may be compressed by tumor cells despite an increase in the number of vessels, thereby forming non-functional channels [25, 26]. Therefore, the effective blood pool available for contrast medium retention is reduced, although tumors with high PD-L1 expression contain a higher density of microvessels. This may explain the significantly lower A-λHU in the high PD‑L1 expression group than in the intermediate- and negative-expression groups. During the arterial phase of contrast‑enhanced scanning, intratumoral iodine concentration primarily reflects the effective blood volume per unit tissue volume, which depends on both vascular density and vascular caliber and patency. High PD‑L1 expression may be associated with increased tumor vascularity; however, structural abnormalities of neovessels reduce intratumoral iodine concentration. Conversely, venous-phase hemodynamics primarily reflect tumor microcirculation and interstitial contrast diffusion and washout, representing tumor neovascular permeability. Increased vascular permeability accelerates contrast washout and lowered intratumoral iodine concentration in the venous phase [23]. This may account for the lack of a significant association between venous-phase PCCT spectral parameters and PD‑L1 expression.
Multivariate logistic regression analysis revealed that A-40 keV was predictive of high PD-L1 expression, which may be attributed to the following reasons. From a physical perspective, the X-ray absorption K-edge of iodine is approximately 33.2 keV [27]. The photoelectric absorption efficiency of iodine for X-rays reaches a peak near this energy level, as 40 keV is close to the K-edge of iodine, enabling highly sensitive detection of subtle changes in intratumoral iodine concentration [27, 28]. Conversely, 70 and 100 keV represent intermediate–high photon energies, distant from the K-edge of iodine, at which the absorption efficiency of iodine is markedly reduced, leading to a significantly diminished ability to detect changes in iodine concentration [29]. Accordingly, the high sensitivity of 40 keV to iodine enables the amplification of minor differences in intravascular iodine retention during the arterial phase into robust changes in CT attenuation, thereby facilitating the precise identification of subtle disparities between the PD-L1 high-expression group and the negative- and moderate-expression groups. This capability is not achievable at higher energy levels.
ROC curve analysis revealed that A-40 keV and A-λHU generated favorable predictive performance for high PD-L1 expression, with AUCs of 0.757 and 0.732, respectively (P < 0.001), indicating good discriminatory ability, which is consistent with the partial results reported by Chen ML et al. regarding the prediction of PD-L1 expression in lung adenocarcinoma using conventional spectral CT [30]. In the present study, further analysis was conducted on the predictive capability of PCCT parameters, clinical characteristics, and their combination, demonstrating that all three models could predict high PD-L1 expression, with the combined model exhibiting the best performance. It helps identify patient subgroups with high PD-L1 expression, thereby guiding individualized immunotherapy decisions for these patients.
Several limitations of this study should be acknowledged. First, A‑λHU, as the slope of the spectral attenuation curve, may be influenced not only by iodine concentration but also by other intratumoral matrix components such as hemorrhage, mucus, and calcification. Although we excluded cases with obvious massive hemorrhage, extensive calcification, or mucus‑rich lesions at enrollment based on PCCT findings, and carefully avoided residual abnormal components during ROI delineation, the potential interference of microscopic calcification, hemorrhage, or mucus that is not visible to the naked eye cannot be completely excluded. Second, tumor necrosis and cystic changes can significantly reduce CT attenuation values, and the 40 keV monochromatic sequence is highly sensitive to iodine concentration, making the interference from such tissues particularly prominent. While we deliberately avoided grossly visible necrotic and cystic areas during ROI placement, we did not quantitatively assess the proportion of necrosis or cystic components across different PD‑L1 expression subgroups, and microscopic necrotic foci may still affect the measurements via partial volume effects. Third, the complete absence of female patients in the PD‑L1 high‑expression group (0/13) led to complete separation in the multinomial logistic regression and limits the generalizability of our findings. Although we performed a sensitivity analysis by removing the gender variable, and the imbalance partly reflects the male predominance of squamous cell carcinoma and potential sex‑based biological differences, institutional selection bias cannot be entirely excluded. Fourth, the small sample size (n = 63, with only 13 high‑expression cases) may affect the stability of the results, and no external validation was performed. However, we performed internal validation using 1,000 bootstrap resamplings. The optimism-corrected AUC of the combined model decreased from 0.889 to 0.847 (95% CI: 0.778–0.949), indicating that the model has reasonable robustness. Future studies with larger sample sizes, gender‑balanced cohorts, and histopathological correlation are needed to further validate the clinical value of A‑λHU and A‑40 keV as imaging biomarkers.
Conclusion
This study demonstrates that arterial-phase iodine uptake parameters from PCCT, especially A-λHU, represent a noninvasive, accurate, and quantitative imaging biomarker. It effectively identifies pretreatment patients with NSCLC with high PD-L1 expression who may benefit from immunotherapy, thereby providing important evidence for personalized treatment planning and addressing the limitations of invasive biopsy.
Supplementary Information
Below is the link to the electronic supplementary material.
Author contributions
JR and LL came up with the design and conception. The PCCT imaging analysis was performed by ZF and JR. The data statistic was conducted by JR and LL. The original writing of the draft and its editing were by YH and YY. All authors contributed to the study, reviewed the data analysis, and approved the final version of the manuscript.
Funding
This research was supported by the National Natural Science Foundation of China (Grant Nos. 12275162 and 12575365).
Data availability
The datasets generated and analyzed during the current study are not publicly available due to institutional and ethical restrictions on the disclosure of individual patient-level data, but are available from the corresponding author on reasonable request, subject to institutional review board approval.
Declarations
Ethics approval and consent to participate
This study was performed in line with the principles of the Declaration of Helsinki. This study was approved by the Ethics Committee of Shandong Cancer Hospital. Written informed consent was waived due to the retrospective design of this study.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Yong Huang, Email: huangyong1970888@sina.com.
Yong Yin, Email: yinyong@sd-cancer.com.
References
- 1.Bray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. Cancer J Clin. 2024;74(3):229–63. [DOI] [PubMed] [Google Scholar]
- 2.Zhao G, Zhang H, Xu F, Lu C, Zhu Q, Grossi F, et al. Neoadjuvant pembrolizumab and chemotherapy in resectable clinical stage III non-small-cell lung cancer: a retrospective cohort study. Translational Lung Cancer Res. 2023;12(1):141–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Leighl NB, Hellmann MD, Hui R, Carcereny E, Felip E, Ahn MJ, et al. Pembrolizumab in patients with advanced non-small-cell lung cancer (KEYNOTE-001): 3-year results from an open-label, phase 1 study. Lancet Respiratory Med. 2019;7(4):347–57. [DOI] [PubMed] [Google Scholar]
- 4.National Comprehensive Cancer Network. NCCN Clinical Practice Guidelines in Oncology: Non-Small Cell Lung Cancer. Version 3.2026. National Comprehensive Cancer Network; December 24, 2025. Accessed April 13, 2026.
- 5.Zhao X, Zhao Y, Zhang J, Zhang Z, Liu L, Zhao X. Predicting PD-L1 expression status in patients with non-small cell lung cancer using [18F] FDG PET/CT radiomics. EJNMMI Res. 2023;13(1). [DOI] [PMC free article] [PubMed]
- 6.Wen Q, Yang Z, Dai H, Feng A, Li Q. Radiomics study for predicting the expression of PD-L1 and tumor mutation burden in non-small cell lung cancer based on CT images and clinicopathological features. Front Oncol. 2021;11. [DOI] [PMC free article] [PubMed]
- 7.Li B, Su J, Liu K, Hu C. Deep learning radiomics model based on PET/CT predicts PD-L1 expression in non-small cell lung cancer. Eur J Radiol Open. 2024;12. [DOI] [PMC free article] [PubMed]
- 8.Liu X, Chen XL, Yuan Y, Pu H, Li H. Dual-energy CT quantitative parameters for prediction of prognosis in patients with resectable rectal cancer. Eur Radiol. 2025;35(8):4945–56. [DOI] [PubMed] [Google Scholar]
- 9.Zhang R, Zheng H, Lin J, Wang J. Review of the application of dual-energy CT combined with radiomics in the diagnosis and analysis of lung cancer. J Appl Clin Med Phys. 2025;26(4). [DOI] [PMC free article] [PubMed]
- 10.Lin LY, Zhang Y, Suo ST, Zhang F, Cheng JJ, Wu HW. Correlation between dual-energy spectral CT imaging parameters and pathological grades of non-small cell lung cancer. Clin Radiol. 2018;73(4):412.e411-412.e417. [DOI] [PubMed]
- 11.Li Q, Li X, Li XY, He XQ, Chu ZG, Luo TY. Histological subtypes of solid-dominant invasive lung adenocarcinoma: differentiation using dual-energy spectral CT. Clin Radiol. 2021;76(1):e7771–7. [DOI] [PubMed] [Google Scholar]
- 12.Bruno E, Palmisano A, Camisassa E, Vignale D, Tacchetti C, Esposito A. Photon-counting detector CT in oncology: a new era of cancer imaging. Insights Imaging. 2026;17(1). [DOI] [PMC free article] [PubMed]
- 13.Zhou Y, Guo X, Lei L, Zhang H, Wang Z, Guo Y, et al. Photon-counting CT versus energy-integrating detector CT performance for various BMI and tumor sizes in lung cancer. Radiology. 2026;318(2). [DOI] [PubMed]
- 14.Zhen KP, Feng Y, Xuan W, Han XA, Jing XW, et al. Clinicopathological and prognostic significance of programmed cell death ligand1 (PD-L1) expression in patients with non-small cell lung cancer: a meta-analysis. J Thorac Dis. 2015;7(3):462–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Wang S, Xu M, Liu Y, Hou X, Gao Z, Sun J, et al. PD-L1 expression and its association with clinicopathological and computed tomography features in surgically resected non-small cell lung cancer: a retrospective cohort study. Sci Rep. 2025;15(1). [DOI] [PMC free article] [PubMed]
- 16.Wu J, Sun W, Wang H, Huang X, Wang X, Jiang W, et al. The correlation and overlaps between PD-L1 expression and classical genomic aberrations in Chinese lung adenocarcinoma patients: a single center case series. Cancer Biology Med. 2019;16(4):811–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Mehta A, Priya VPL, Teja L, Pillai K. The prevalence of programmed death ligand-1 (PD-L1) expression in non-small cell lung cancer: an experience from Tertiary care hospital. Cureus. 2024. [DOI] [PMC free article] [PubMed]
- 18.Ohara S, Suda K, Hamada A, Chiba M, Ito M, Shimoji M, et al. Clinical factors associated with high PD-L1 expression in patients with early‐stage non‐small cell lung cancer. Thorac Cancer. 2024;15(31):2229–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Huang Y, Cho HJ, Stranger BE, Huang RS. Sex dimorphism in response to targeted therapy and immunotherapy in non-small cell lung cancer patients: a narrative review. Translational Lung Cancer Res. 2022;11(5):920–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Sauer ST, Huflage H, Christner SA, Patzer TS, Kiesel M, Quenzer A, et al. Preliminary experience with virtual monoenergetic imaging and iodine mapping in the primary staging of endometrial cancer. Cancers. 2024;16(6). [DOI] [PMC free article] [PubMed]
- 21.Zhao L, Zhou W, Fu Y, Ge Y, Feng L, Wang X, et al. Diagnostic value of one-stop CT energy spectrum and perfusion for angiogenesis in colon and rectum cancer. BMC Med Imaging. 2024;24(1). [DOI] [PMC free article] [PubMed]
- 22.Spira D, Neumeister H, Spira SM, Hetzel J, Spengler W, et al. Assessment of tumor vascularity in lung cancer using volume perfusion CT (VPCT) with histopathologic comparison: a further step toward an individualized tumor characterization. J Comput Assist Tomogr. 2013;37(1):15–21. [DOI] [PubMed] [Google Scholar]
- 23.Yu ZH, Ma WQ, Ren JW, Zhang XT, Chu L. Role of Computed Tomography in Predicting Programmed Death Ligand-1 Positivity in Gastric Adenocarcinoma. J Multidisciplinary Healthc. 2025;18:609–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Bichsel CA, Wang L, Froment L, Berezowska S, Müller S, Dorn P. Increased PD-L1 expression and IL-6 secretion characterize human lung tumor-derived perivascular-like cells that promote vascular leakage in a perfusable microvasculature model. Sci Rep. 2017;7(1). [DOI] [PMC free article] [PubMed]
- 25.Pries AR, Höpfner M, le Noble F, Dewhirst MW, Secomb TW. The shunt problem: control of functional shunting in normal and tumour vasculature. Nat Rev Cancer. 2010;10(8):587–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Fukumura D, Jain RK. Tumor microvasculature and microenvironment: Targets for anti-angiogenesis and normalization. Microvasc Res. 2007;74(2–3):72–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Nagayama Y, Tanoue S, Inoue T, Oda S, Nakaura T, Utsunomiya D, et al. Dual-layer spectral CT improves image quality of multiphasic pancreas CT in patients with pancreatic ductal adenocarcinoma. Eur Radiol. 2019;30(1):394–403. [DOI] [PubMed] [Google Scholar]
- 28.Fahrni G, Boccalini S, Lacombe H, Oliveira F, Houmeau A, Francart F, et al. Ultra-high-resolution 40 keV virtual monoenergetic imaging using spectral photon-counting CT in high-risk patients for coronary stenoses. Eur Radiol. 2024;35(6):3042–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Yamauchi H, Buehler M, Goodsitt MM, Keshavarzi N, Srinivasan A. Dual-Energy CT–Based Differentiation of Benign Posttreatment Changes from Primary or Recurrent Malignancy of the Head and Neck: Comparison of Spectral Hounsfield Units at 40 and 70 keV and Iodine Concentration. Am J Roentgenol. 2016;206(3):580–7. [DOI] [PubMed] [Google Scholar]
- 30.Chen ML, Shi AH, Li Xt, Wei YY, Qi LP, Sun YS. Is there any correlation between spectral CT imaging parameters and PD-L1 expression of lung adenocarcinoma? Thorac Cancer. 2019;11(2):362–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets generated and analyzed during the current study are not publicly available due to institutional and ethical restrictions on the disclosure of individual patient-level data, but are available from the corresponding author on reasonable request, subject to institutional review board approval.