Decision-making tree model for distinguishing between benign and malignant subpleural pulmonary lesions based on ultrasonic strain elastography

Abstract

Objectives

To evaluate ultrasonic strain elastography (USE) findings of subpleural pulmonary lesions (SPLs) and generate a decision-making tree model to distinguish between benign and malignant SPLs.

Methods

We prospectively analyzed the USE findings in patients with SPLs from May 2020 to June 2021. The findings included the strain ratio of the lesion (including non-perfusion areas) to the chest wall muscle (referred to as SRI), the strain ratio of the lesion (excluding non-perfusion areas) to the chest wall muscle (referred to as SRE), and the water ripple sign. The patients were divided into a development cohort (DC) and a validation cohort (VC). Using the DC, we employed conditional inference tree analysis to generate a decision-making tree model based on the USE findings. The diagnostic performances of the models were evaluated using the VC.

Results

In total, 353 patients were eligible: 247 in the DC and 106 in the VC group. The SRI, SRE, and water ripple sign were good discriminators between benign and malignant SPLs. The model employing the SRI (with a cut-off of 11.6) showed 92.16% sensitivity, 81.82% specificity and 86.79% accuracy, whereas the model employing the SRE (with a cut-off of 11.6) achieved 94.12% sensitivity, 85.46% specificity and 89.62% accuracy.

Conclusions

The decision-making tree model based on USE effectively distinguished between benign and malignant SPLs. We suggest that the SRI or SRE be evaluated first and the water ripple sign should be evaluated subsequently in cases where the SRI or SRE exceeds 11.6.