Breast lesion shape and margin evaluation: BI-RADS based metrics understate radiologists' actual levels of agreement. (1st May 2018)

• Record Type: Journal Article
• Authors: Rawashdeh, Mohammad; Lewis, Sarah; Zaitoun, Maha; Brennan, Patrick
• Is Part Of: Computers in biology and medicine. Volume 96(2018)
• Page Start: 294
• Page End: 298
• Publication Date: 2018-05-01
• Subjects: Lesion delineation -- Radiotherapy -- Breast lesion -- Mammography -- Electronic stylus; Medicine -- Data processing -- Periodicals; Biology -- Data processing -- Periodicals; 610.285
• DOI: 10.1016/j.compbiomed.2018.04.005
• Languages: English
• ISSNs: 0010-4825

Abstract

Abstract: Background: While there is much literature describing the radiologic detection of breast cancer, there are limited data available on the agreement between experts when delineating and classifying breast lesions. The aim of this work is to measure the level of agreement between expert radiologists when delineating and classifying breast lesions as demonstrated through Breast Imaging Reporting and Data System (BI-RADS) and quantitative shape metrics. Methods: Forty mammographic images, each containing a single lesion, were presented to nine expert breast radiologists using a high specification interactive digital drawing tablet with stylus. Each reader was asked to manually delineate the breast masses using the tablet and stylus and then visually classify the lesion according to the American College of Radiology (ACR) BI-RADS lexicon. The delineated lesion compactness and elongation were computed using Matlab software. Intraclass Correlation Coefficient (ICC) and Cohen's kappa were used to assess inter-observer agreement for delineation and classification outcomes, respectively. Results: Inter-observer agreement was fair for BI-RADS shape (kappa = 0.37) and moderate for margin (kappa = 0.58) assessments. Agreement for quantitative shape metrics was good for lesion elongation (ICC = 0.82) and excellent for compactness (ICC = 0.93). Conclusions: Fair to moderate levels of agreement was shown by radiologists for shape and margin classifications of cancers using the BI-RADS lexicon. When quantitative shape metrics were used to evaluate radiologists' delineation of lesions, good to excellent inter-observer agreement was found. The results suggest that qualitative descriptors such as BI-RADS lesion shape and margin understate the actual level of expert radiologist agreement. Highlights: Fair levels of agreement were found when readers classified lesions using BI-RADS. Using computer metrics, good agreement was found for lesion delineation. Further work is needed to explore how radiologists apply the BI-RADS criteria.