Objectives To determine the significance of density homogeneity in differentiating malignant and benign subcentimeter solid nodules (SNs). Materials and methods Between January 2018 and July 2024, 735 subcentimeter malignant SNs (SMSNs) and 814 subcentimeter benign SNs (SBSNs) from Center 1 (training set), as well as 244 SBSNs and 163 SMSNs from two other centers (validation set), were included. Patients’ clinical characteristics (e.g., age, gender) and CT features of lesions (e.g., density homogeneity, diameter) were analyzed and compared, focusing on assessing the significance of density homogeneity in differential diagnosis. The optimal cutoff value of standard deviation (SD) of mean CT value for distinguishing visually heterogeneous and homogeneous SNs was calculated and validated, respectively. Results The optimal SD cutoff for density heterogeneity was 57.3 HU, with an area under the curve (AUC) of 0.928 (p < 0.001) in the validation set. In training set, heterogeneous density demonstrated the highest predictive efficiency among all CT features (AUC: 0.722), and significantly improved the performance of predictive model after incorporating this indicator (AUC from 0.698 to 0.824, p < 0.001). The model indicated that heterogeneous density, lobulation, spiculation, and air bronchogram sign were independent predictors of SMSNs (all p < 0.05). In validation set, density heterogeneity also demonstrated the highest predictive efficiency (AUC: 0.738), and significantly enhanced the predictive model’s performance (AUC from 0.679 to 0.831, p < 0.001). Conclusions Heterogeneous subcentimeter SNs, especially those with lobulation, spiculation, or air bronchogram sign, should raise a high suspicion of malignancy; therefore, close monitoring is required. Critical relevance statement Accurately differentiating malignant and benign solid pulmonary nodules at the subcentimeter stage is still challenging; the present study confirmed that heterogeneous density of lesions is a stable and reliable radiological predictor of malignancy. Key Points <UnorderedList Mark="Bullet"> <ItemContent> Benign and malignant subcentimeter solid pulmonary nodules share considerable overlaps in CT morphological features, making differential diagnosis challenging. </ItemContent> <ItemContent> Heterogeneous lesions were significantly more common in SMSNs than in benign ones, particularly in the group of nodules ≤ 8 mm. </ItemContent> <ItemContent> Heterogeneous density exhibited superior efficiency than lobulation, spiculation, pleural indentation, vacuole sign and other morphological features in differentiating subcentimeter SNs. </ItemContent> <ItemContent> Heterogeneous subcentimeter SNs, particularly those with lobulation, spiculation, or air bronchogram sign, should be highly suspected for malignancy. </ItemContent> </UnorderedList> Graphical Abstract

Density homogeneity as a crucial CT indicator for differentiating malignant and benign subcentimeter solid pulmonary nodules: A retrospective multi-center study

期刊论文
出版模式：

开放获取
发表日期：

2026年5月16日

查看全文

Wen-tao Zhang,
Hui Gan,
Wei Luo,
Min Zhao,
Can Ding,
Xue-feng Jiang,
Ting Fu,
Fa-jin Lv,
Zhi-gang Chu

摘要

Objectives

To determine the significance of density homogeneity in differentiating malignant and benign subcentimeter solid nodules (SNs).

Materials and methods

Between January 2018 and July 2024, 735 subcentimeter malignant SNs (SMSNs) and 814 subcentimeter benign SNs (SBSNs) from Center 1 (training set), as well as 244 SBSNs and 163 SMSNs from two other centers (validation set), were included. Patients’ clinical characteristics (e.g., age, gender) and CT features of lesions (e.g., density homogeneity, diameter) were analyzed and compared, focusing on assessing the significance of density homogeneity in differential diagnosis. The optimal cutoff value of standard deviation (SD) of mean CT value for distinguishing visually heterogeneous and homogeneous SNs was calculated and validated, respectively.

Results

The optimal SD cutoff for density heterogeneity was 57.3 HU, with an area under the curve (AUC) of 0.928 (p < 0.001) in the validation set. In training set, heterogeneous density demonstrated the highest predictive efficiency among all CT features (AUC: 0.722), and significantly improved the performance of predictive model after incorporating this indicator (AUC from 0.698 to 0.824, p < 0.001). The model indicated that heterogeneous density, lobulation, spiculation, and air bronchogram sign were independent predictors of SMSNs (all p < 0.05). In validation set, density heterogeneity also demonstrated the highest predictive efficiency (AUC: 0.738), and significantly enhanced the predictive model’s performance (AUC from 0.679 to 0.831, p < 0.001).

Conclusions

Heterogeneous subcentimeter SNs, especially those with lobulation, spiculation, or air bronchogram sign, should raise a high suspicion of malignancy; therefore, close monitoring is required.

Critical relevance statement

Accurately differentiating malignant and benign solid pulmonary nodules at the subcentimeter stage is still challenging; the present study confirmed that heterogeneous density of lesions is a stable and reliable radiological predictor of malignancy.

Key Points

Benign and malignant subcentimeter solid pulmonary nodules share considerable overlaps in CT morphological features, making differential diagnosis challenging.

Heterogeneous lesions were significantly more common in SMSNs than in benign ones, particularly in the group of nodules ≤ 8 mm.

Heterogeneous density exhibited superior efficiency than lobulation, spiculation, pleural indentation, vacuole sign and other morphological features in differentiating subcentimeter SNs.

Heterogeneous subcentimeter SNs, particularly those with lobulation, spiculation, or air bronchogram sign, should be highly suspected for malignancy.

Graphical Abstract