Purpose To explore the added value of quantitative parameters derived from routine chest contrast-enhanced computed tomography (CECT) in distinguishing malignant from benign solid pulmonary nodules (SPNs). Methods Eighty-one SPNs pathologically confirmed as benign or malignant with preoperative nonenhanced chest and CECT scans were retrospectively analyzed. Quantitative parameters [CT attenuation value of nonenhanced phase (AVN), value of arterial phase (AVA), value of venous phase (AVV), their differentials (△AVA−N, △AVV−N, △AVV−A), diameter] and qualitative CT features [lobulation, spiculation, vacuolar sign, pleural depression sign, vascular convergence, edge clarity] were obtained. Inter-group comparisons for clinical/imaging variables used t-tests/Mann-Whitney U tests or Chi-square/Fisher’s tests. Three multivariate logistic regression models (qualitative, quantitative, and combined models) were developed and evaluated through five-fold cross-validation, DeLong tests (Bonferroni-corrected α = 0.0167), decision/calibration curves, and Bootstrap-based threshold sensitivity analysis (1000 iterations; 0.1–0.9 thresholds). Subgroup ROC analyses assessed age/diameter effects (stratified by mean/median). Results Malignant SPNs showed higher age and greater AVA, AVN, AVV, △AVA−N, △AVV−N, diameter (P < 0.05). AVV showed the strongest discriminatory power among quantitative parameters (AUC = 0.779). The qualitative model incorporated vascular convergence, pleural depression sign, and lobulation, while the quantitative model included AVV. Cross-validation yielded mean AUCs of 0.877 ± 0.019, 0.790 ± 0.081, and 0.900 ± 0.042 for the qualitative, quantitative, and combined model respectively. The combined model surpassed the qualitative model (P = 0.016), demonstrating better calibration and decision curve performance. Bootstrap analysis identified 0.4 as the optimal sensitivity-specificity threshold. Subgroup AUCs were 0.945/0.860 (mean-age strara) and 0.903/0.912 (median-diameter strata). Conclusions Quantitative CECT parameters, particularly AVV, aid in discriminating malignant SPNs. Combining AVV with qualitative features enhances diagnostic accuracy for malignancy risk assessment.

The added value of quantitative contrast-enhanced CT parameters in distinguishing malignant from benign solid pulmonary nodules

期刊论文
出版模式：

开放获取
发表日期：

2025年7月24日

查看全文

Peng Yun,
Zhencheng Zhan,
Zijing Wu,
Yutong Rao,
Ke Sun,
Lianggeng Gong

摘要

Purpose

To explore the added value of quantitative parameters derived from routine chest contrast-enhanced computed tomography (CECT) in distinguishing malignant from benign solid pulmonary nodules (SPNs).

Methods

Eighty-one SPNs pathologically confirmed as benign or malignant with preoperative nonenhanced chest and CECT scans were retrospectively analyzed. Quantitative parameters [CT attenuation value of nonenhanced phase (AV_N), value of arterial phase (AV_A), value of venous phase (AV_V), their differentials (△AV_A−N, △AV_V−N, △AV_V−A), diameter] and qualitative CT features [lobulation, spiculation, vacuolar sign, pleural depression sign, vascular convergence, edge clarity] were obtained. Inter-group comparisons for clinical/imaging variables used t-tests/Mann-Whitney U tests or Chi-square/Fisher’s tests. Three multivariate logistic regression models (qualitative, quantitative, and combined models) were developed and evaluated through five-fold cross-validation, DeLong tests (Bonferroni-corrected α = 0.0167), decision/calibration curves, and Bootstrap-based threshold sensitivity analysis (1000 iterations; 0.1–0.9 thresholds). Subgroup ROC analyses assessed age/diameter effects (stratified by mean/median).

Results

Malignant SPNs showed higher age and greater AV_A, AV_N, AV_V, △AV_A−N, △AV_V−N, diameter (P < 0.05). AV_V showed the strongest discriminatory power among quantitative parameters (AUC = 0.779). The qualitative model incorporated vascular convergence, pleural depression sign, and lobulation, while the quantitative model included AV_V. Cross-validation yielded mean AUCs of 0.877 ± 0.019, 0.790 ± 0.081, and 0.900 ± 0.042 for the qualitative, quantitative, and combined model respectively. The combined model surpassed the qualitative model (P = 0.016), demonstrating better calibration and decision curve performance. Bootstrap analysis identified 0.4 as the optimal sensitivity-specificity threshold. Subgroup AUCs were 0.945/0.860 (mean-age strara) and 0.903/0.912 (median-diameter strata).

Conclusions

Quantitative CECT parameters, particularly AV_V, aid in discriminating malignant SPNs. Combining AV_V with qualitative features enhances diagnostic accuracy for malignancy risk assessment.