Can Density Homogeneity on Chest CT Improve Differentiation of Sub-Centimeter Nodules?

Emerging research suggests that density homogeneity may be a more viable differentiator than traditional computed tomography (CT) features for identifying sub-centimeter malignant solid lung nodules.

For the retrospective study, recently published in Insights into Imaging, researchers reviewed chest CT data for over 1,900 sub-centimeter solid lung nodules to determine the prognostic capacity of a predictive model and key features on CT in differentiating malignant and benign nodules. The training set data included 735 sub-centimeter malignant solid nodules (SMSNs) and 814 sub-centimeter benign solid nodules (SBSNs), and the external validation set was comprised of 163 SMSNs and 244 SBSNs, according to the study.

For the differentiation of SMSNs and SBSNs in the training set, the study authors found that density homogeneity had a 72.2 percent AUC in contrast to 57.2 percent for lobulation, 55 percent for nodule margin, 54.9 percent for pleural indentation and 54.3 percent for the halo sign.

In validation testing, incorporating density heterogeneity into a predictive model improved the model’s AUC by more than 15 percent (67.9 percent vs. 83.1 percent).

“In contrast to clinical characteristics and traditional CT features, density homogeneity demonstrated substantial diagnostic efficiency, which may be associated with the histopathological heterogeneity between SMSNs and SBSNs,” noted lead study author Wen-tao Zhang, MD, who is affiliated with the Department of Radiology at the First Affiliated Hospital of Chongqing Medical University in Chongqing, China, and colleagues.

The researchers found that heterogeneous SMSNs had more than double the percentage of adenocarcinomas in situ (23.88 percent vs. 10.17 percent) and more than triple the number of minimally invasive adenocarcinomas (30.45 percent vs. 9.89 percent) in comparison to homogeneous SMSNs.

Homogenous SMSNs had significantly higher percentages of invasive adenocarcinomas (65.54 percent vs. 40.68 percent) and squamous cell carcinoma (5.93 percent vs. 1.05 percent) in contrast to heterogeneous SMSNs, according to the study authors.

“ … It is possible that the density of heterogeneous SMSN may progressively increase with the advancement of invasive pathological components progressing and fibrous tissue proliferation, ultimately resulting in a radiologically homogeneous appearance. In addition, the mixed distribution of invasive components, mucus, fibrous tissue, or residual airspace may also be associated with heterogeneous density on CT images,” posited Zhang and colleagues.

(Editor’s note: For related content, see “FDA Issues 510(k) Clearance of AI-Powered Assessment for Lung Cancer on Low-Dose CT Scans,” “Study Looks at Combining PCCT and Lung Texture Analysis for Evaluating ILD in Patients with Systemic Sclerosis” and “Study Shows Merits of CT Vascular Sign for Differentiating Solid Pulmonary Nodules.”)

In regard to study limitations, the authors acknowledged that minor scanning variations for CT scanners may impact the assessment of density homogeneity. The researchers also conceded that the lack of evaluation of nodules through a mediastinal window may lead to classification of some heterogeneous solid nodules as sub-solid nodules.