The values of elastic quantitative and semi-quantitative indexes measured from different frequencies in the establishment of prediction models for breast tumor diagnosis

Abstract

Objective: To obtain the elastic quantitative and semi-quantitative indexes of solid breast masses using ultrasound linear array probes with two different frequencies, and to construct prediction models and evaluate their diagnostic values.

Methods: A total of 92 patients who were scheduled for surgical treatment on solid breast masses were enrolled in this study. Linear array probes with two frequencies, 9-3 MHz (L9 group) and 14-5 MHz (L14 group), were used for sound touch elastography and strain elastography before surgery, and the maximum elasticity value (Emax), average elasticity value (Emean), minimum elasticity value (Emin), standard deviation (SD)(in kPa), elasticity ratio (E), and strain ratio to fat (SRf) were recorded and calculated for the breast mass (A) and surrounding tissues (Shell). The elastic characteristic indexes of the L9 group and L14 group were compared, and the prediction models of these two groups were constructed using Logistic regression method.

Results: The diagnostic performance of the prediction model based on L9 group was better than the model based on L14 group (AUC: 0.904 vs. 0.810, P = 0.0343, z = 2.116) and the best single index EMax-shell-L9 (P = 0.0398, z = 2.056). The sensitivity of L9 based model was 85.19% and the specificity was 84.21%.

Conclusion: The prediction model based on quantitative and semi-quantitative elastic ultrasound indexes from L9-3 probe exhibited better performance, which could improve the diagnostic accuracy for malignant breast tumors.

Keywords: Breast; Elastography; Frequency; Prediction model; Ultrasound.