Abstract
Purpose: Breast Imaging Reporting and Data System (BI-RADS) has many descriptors whose combined evaluation leads to the differentiation of benignancy and malignancy of breast lesions. The purpose of this study was to determine the best combination.
Materials and Methods: Institutional Review Board approval was waived for this retrospective analysis. Five radiologists evaluated BI-RADS descriptors of 109 breast lesions of 71 consecutive patients (mean age 45.3 years, range 22-70 years) and 5 descriptors (shape, orientation, margin, echo pattern, posterior features) were collected as binary type data. Using SPSS package, for each radiologist, cross tabulation analysis was conducted between each of the descriptors and pathologic diagnosis. Subsequently, logistic regression analysis was conducted for variable combinations of descriptors, and the best model was selected. Finally, the same step was used for the dataset made by 5 radiologists together.
Results: Sensitivity and specificity of 5 descriptors were evaluated by 5 radiologists. Shape showed 76% of sensitivity (56-92%) and 52% of specificity (39-64%). Orientation showed 51% of sensitivity (31-72%) and 83% of specificity (73-94%). Margin showed 83% of sensitivity (72-90%) and 53% of specificity (36-71%). Echo pattern showed 87% of sensitivity (54-100%) and 22% of specificity (3-60%). Posterior features showed 34% of sensitivity (21-46%) and 86% of specificity (71- 97%). As a result, relatively high sensitive descriptor was echo pattern and relatively high specific descriptor was posterior feature.
Subsequently, regression model for the large dataset output by 5 radiologists together were made by ≥ 2 descriptors and the best model that showed the highest PCC (Percentage Correctly Classified) was shape + orientation (72.1, range 67.9-72.5) in combination of two. In combination of three, the best model was shape + orientation + posterior feature (72.5, range 71.6-76.1). Shape + orientation+ posterior feature + echo pattern (73.8, range 71.6-77.1) was the best model for combination of four and in combination of five, the PCC was 72.7(range 72.5-76.1).
Conclusion: The highest PCC in each of the combinations showed no significant difference. Therefore, combination of two descriptors was sufficient for regression model and the combination of shape and orientation was the most significant descriptors in distinction of benign and malignant breast masses.
Keywords: Breast neoplasm, BI-RADS lexicon, regression model, ultrasonograpy, lesions, PCC.
Graphical Abstract
Call for Papers in Thematic Issues
Intelligent Information Retrieval for Multispectral and Hyperspectral Imaging
Multispectral and hyperspectral imaging is an interdisciplinary research area, widely focuses on spectral, spatial, and temporal data. It offers a plethora of opportunities to efficiently analyze the vast area of the earth's surface. However, there are numerous factors such as dimensionality and size of the hyperspectral data, lesser training samples, ...read more
Machine and Deep Learning in Medical Image Diagnosis
Medical Image Processing regards a set of methodologies that have been developed over recent years with the purpose of improving medical image quality, improving medical data visualization, understanding, and assisting medical diagnosis, and so on. The application of machine and deep learning methods in sensing and imaging can potentially have ...read more
Quantum Machine Learning for Medical Data and Imaging
Quantum computing has promised a significant speedup in certain computationally intensive tasks that are intractable on classical computers. Researchers in quantum machine learning, such as those working in computer vision, image processing, biomedical analysis, and related topics, may play an important role in comprehending and working on complicated medical data, ...read more
19
4
1
1