Comparison of the Diagnostic Performance of Breast Ultrasound and CAD Using BI-RADS Descriptors and Quantitative Variables

Abstract

Background: Computer-aided diagnosis (CAD) as an adjunct to breast ultrasonography (US) has been extensively developed using different breast imaging techniques to help improve radiologists' diagnostic performance, particularly in distinguishing malignant from benign breast lesions. Additionally, quantitative variables can be obtained using CAD. Objectives: The aim of this study was to evaluate the usefulness of CAD as an adjunct to breast US by exploring the diagnostic performance of US and CAD using the breast imaging reporting and data system (BI-RADS) descriptors and quantitative variables. Patients and Methods: Between October 2015 and December 2016, we performed breast US with CAD (S-Detect) for screening or diagnostic purposes. We assessed the diagnostic performance according to the BI-RADS descriptors, quantitative variables (width, height, height/width (H/W) ratio, area, and depth), and combined results. We analyzed the agreement in the descriptors between US and CAD. Results: Four hundred thirty-two women with 521 breast lesions were enrolled. The area under the receiver operating characteristic curve (AUC), sensitivity, and specificity of US vs. CAD were 0.82 vs. 0.78, 95% vs. 78%, and 69% vs. 78%, respectively. When CAD was combined with breast US, the specificity was significantly improved from 69% to 72% (P < 0.05). Among all descriptors and quantitative variables, the height and H/W ratio exhibited the greatest AUC (0.76 and 0.75), and the H/W ratio had the highest sensitivity (90%). According to the concordance analysis, the orientation exhibited the greatest agreement (k = 0.57). Conclusion: Using CAD, we could accurately determine the BI-RADS descriptors, quantitative variables and improve the specificity. Additionally, the orientation and H/W ratio are consistent key factors for both US and CAD.

Ultrasonographic morphological characteristics determined using a deep learning-based computer-aided diagnostic system of breast cancer

“…[ 24 , 26 , 56 , 57 ] However, to the best of our knowledge, no study has analyzed raw data itself, driven by DL-CAD, to quantify the morphological characteristics of breast cancers. Kim et al [ 58 ] analyzed the diagnostic performance of breast US using quantitative variables, but the quantitative variables in that study were width, height, height/width ratio, area, and depth. This is the first study focusing on the correlation between molecular subtype and sonographic features of breast cancer using quantitative analysis by DL-CAD with breast US.…”

Barriers and facilitators of artificial intelligence conception and implementation for breast imaging diagnosis in clinical practice: a scoping review

“…MRI studies especially lacked data (from 93 [59] to 1715 patients [25] in included studies, with varying types of MR protocols). While US imaging may benefit from larger accessibility, an important discrepancy was observed among studies where the number of patients ranged from 92 to 5151 [54,[60][61][62][63][64][65][66][67]. Variety in the data was also reported as a critical issue since clinical data is often imbalanced due to disease prevalence, data availability, or population characteristics, which remains an unsolved issue still under research in the ML field [16].…”