Eur Radiol: How to layer the risk of malignant breast non-lump lesions?

The Mammography Reporting and Data System (BI-RADS) defines a lump as a three-dimensional occupied lesions observed on two different planes that can be distinguished from normal anatomical structures.
However, high-resolution breast ultrasound (US) is often found to show limited asymmetrical, no visible edges or shapes on two orthocond planes, and therefore does not meet the strict criteria for BI-RADS definition of "lumps".
Since BI-RADS in the United States has not yet defined such lesions, radiologists define them as non-lump-like lesions (breast-non-mass lesions, NMLs) or non-block-like lesions corresponding to non-block-like enhancement on mammography (MRI).
accurate definition of breast NMLs is of clinical significance, as it can be manifested as NMLs from malignant catheter in-place cancer to benign fibrocystic changes.
although some studies have been conducted on breast NMLs, there is still a lack of understanding of these lesions. At this stage, there are differences in the identification and description of breast NMLs by ultrasound doctors, and no standard guidelines for the interpretation and treatment of breast NMLs have been established.
recently, a study published in the journal European Radiology established a classification system that uses imaging characteristics to explain non-lump-like lesions (NMLs) found on the US and stratify their cancer risk.
review included 715 breast cancer patients from 2012 to 2016, with a total of 715 breast NMLs.
each patient was diagnosed with a mammography test.
radiologists evaluated mammary US and target characteristics and the final BI-RADS classification.
using the multivarivarial logic regression method to look for image features associated with malignantity in the development dataset (n s 460).
based on the advantage ratio (ORs) of imaging characteristics significantly related to malignantity, a system for classifying BI-RADS categories (3 to 5) was developed and validated in different validation data set (n s 255).
715 NMLs, 385 (53.8%) were benign and 330 (46.2%) were malignant.
in the development data set, the following US characteristics are associated with malignancy (all p .lt; 0.001): segment distribution (OR s 3.03; 95% confidence interval (CI), 1.50-6.15), related calcification (OR s 4.26; 95% CI, 1.62 to 11.18), abnormal changes in catheters (OR - 4.91; 95% CI, 2.07-11.68), Rear Sound and Shadow (OR s 20.20; 95% confidence interval, 6.46 - -63.23).
the following mammary target characteristics are also associated with malignancy (average p . . . . 95% CI, 3.06-20.81) and bureau asymmetry (OR s 4.75; 95% confidence interval, 1.90 - -11.88).
in the validation data set, the classification system using mammary US and target showed a larger sub-curve area (0.908-0911) than when the BI-RADS category was not applied for malignant prediction (0.98-0911).
Figure 1 box diagram shows the fitted malignant probability of each BI-RADS category determined by the newly developed system, (a) the US feature plus target characteristics, and (b) using the newly developed classification system and the ROC curve of the BI-RADS classification using the newly developed classification system and the radiologist who did not use the new system.
(a) The AAUC of the radiologist (reader 3-4, 0.951-0956) using the US feature plus target characteristic system is higher than that of the unused radiologist (reader 1-2, 0.908-0.911) (p .
(b) The AAU (reader 3 to 4, 0.937 to 0.947) of radiologists using US-only systems is significantly higher than that of two unused radiologists (reader 1 to 2, 0.867 to 0.989) (p .
The newly developed classification system in this study, which combines the US and target characteristics of breast NMLs, can layer cancer risk according to BI-RADS categories, thus helping to clinically interpret and manage all NMLs detected on mammary US, providing technical support for further consensus.