Edited by:
Reviewed by: Rajendra A. Badwe, Tata Memorial Hospital, India
*Correspondence: Hai-Liang Li,
†These authors have contributed equally to this work and share first authorship
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
The application performance of the Breast Imaging-Reporting and Data System (BI-RADS) in primary healthcare settings remains uncertain. The normativity of BI-RADS classification and the efficacy of breast cancer detection guided by BI-RADS classification were evaluated here.
All data used in the current study were derived from a breast cancer screening cohort baseline database, which consists of 8,996 women aged 35–64 years from Central China. Participants aged 35–44 underwent automated breast ultrasound (ABUS) and handheld ultrasound (HHUS), while those aged 45–64 were screened with ABUS, HHUS, and mammography (MG). All imaging diagnoses were made by radiologists according to the BI-RADS 5th edition classification system published by the ACR in 2013. The distribution of malignant imaging findings and inter-modality agreement on BI-RADS classifications were assessed. Based on pathological results, the area under the receiver operating characteristic (ROC) curve (AUC) was used to assess the performance of breast cancer screening according to BI-RADS-guided referrals.
Of individuals found with irregular morphology, 44%, 45.7% and 77.4% were classified as BI-RADS category 4 or higher for HHUS, ABUS and MG, respectively; For those with indistinct margins, the proportion was 81%, 77.5% and 40.8%, correspondingly; For grouped calcifications, they were 100% for HHUS and 85.7% for MG; Meanwhile, 72.7% and 88.9% not parallel (taller than wide) masses were categorized as BI-RADS category 4 for HHUS and ABUS. The concordance of BI-RADS classification was as high as 98.2% between HHUS and ABUS (Kappa = 0.726), whereas it was about 96% between ultrasound and MG (Kappa ranged from 0.21 to 0.25). The BI-RADS guided screening performance for breast cancer showed AUC values of 0.9935 for ABUS, 0.9529 for HHUS, 0.8983 for MG. If the BI-RADS diagnosis of MG was considered in ultrasound-negative women, only the HHUS’s performance was boosted, achieving an AUC of 0.9920.
Radiologists at primary hospitals can effectively apply BI-RADS based on the malignant features they found. BI-RADS can provide a reliable framework for guiding breast cancer screening in primary healthcare settings.
The goal of breast cancer screening is to detect cancer at an early stage, thereby improving treatment outcomes and survival rates (
However, imaging diagnostic accuracy is modulated by radiologists experience (
Despite advancements in imaging technology, the variance in clinical expertise, imaging equipment, and training levels across different levels of hospitals persists, leading to an uncertain efficacy of identifying and interpreting malignant signs and radiological diagnostic capabilities in breast cancer screening within the primary healthcare setting (
Data used in the current study were extracted from a breast cancer screening cohort which consists of 8,996 women aged 35–64 years from Central China. BI-RADS diagnoses, pathological results and imaging characteristics were retrospectively analyzed. Participants enrolled in this screening cohort were local general women aged 35–64 years. The exclusion criteria included being pregnant, lactating, or planning to become pregnancy; had a history of breast tumor resection, contralateral breast surgery, breast augmentation, or percutaneous biopsy within the past 12 months; had a prior tumor diagnosis or treatment within the last 12 months; or exhibited suspicious signs without an imaging indication. This study was approved by the independent ethic committee of Henan Cancer Hospital (Approval Number: 19/109-1893).
In the current study, breast cancer imaging screening modalities included HHUS, ABUS and MG. Based on current screening guidelines (
HHUS was performed using the EADN U50 ultrasound device (frequency range: 7.0-16.0 MHz; EADN, Shenzhen, China), with detailed documentation of breast lesion characteristics. ABUS was operated using the SIUI IBUS BE3 (frequency: 5–12 MHz; Shantou Institute of Ultrasonic Instruments, Shantou, China) and Invenia ABUS (C15-6XW Reverse Curve™, frequency: 6–15 MHz; GE Healthcare, Hatfield, UK) devices for scanning. MG was conducted using the Hologic Selenia Dimensions system (Hologic, Massachusetts, USA). Ultrasound and MG images were interpreted by radiologists according to the BI-RADS 5th edition classification system published by the ACR in 2013.
The key imaging findings serving as the basis for the BI-RADS classification were collected as follows: the size, shape, margin, orientation of the lesion, the presence of calcifications, and other associated signs of suspicious lesions (including architectural distortion, skin thickening, and hardness in elasticity-imaging) for HHUS and ABUS; The size, shape, and density of lesion, the presence of calcifications, and other associated signs (including architectural distortion, skin thickening, and the presence of asymmetries) for MG. Mass shape was categorized as regular(including round, oval) or irregular; Orientation was classified as parallel (long axis of lesion is parallel to the skin, also referred to as wider than tall) or not parallel (long axis of lesion is not parallel to the skin, also referred to as taller than wide); The margin was described as circumscribed or indistinct; Calcifications were classified as grouped, linear or benign-appearing. Architectural distortion was defined as a localized distorted breast parenchyma with no definite mass visible; asymmetries was defined as an asymmetric dense shadow of fibro glandular tissue without a clear three-dimensional outline or distinct margin when compared with the corresponding location on the contralateral breast.
Pathological diagnosis was carried out by pathologists at the screening units in accordance with uniform standards. Women with BI-RADS category 4 or higher underwent biopsy, and confirmed cases were staged according to the eighth-edition Breast Cancer Staging System published by the American Joint Committee on Cancer (AJCC).
Mammography screenings were carried out employing the Hologic Selenia Dimensions system (Hologic, Inc., Marlborough, MA, USA), a system renowned for its high-resolution imaging capabilities, thereby contributing to the accuracy of the diagnostic process.
The distribution of malignant imaging findings across the BI-RADS categories from different imaging screening modalities was calculated to assess the prevalence of high classifications that would warrant clinical biopsy. The inter-modality agreement on BI-RADS category 4 or higher was assessed to evaluate diagnostic characteristics of each imaging method. The efficacy of breast cancer screening based on BI-RADS-guided referrals for pathological biopsy was evaluated using the diagnostic indicators of sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the area under the receiver operating characteristic (ROC) curve. SAS 9.4 software was utilized for all statistical analyses, with a significance level set at α = 0.05.
Of 8996 women, 362 individuals were classified as BI-RADS category 4 or higher, and 117 biopsies were performed in the breast cancer screening program. Ultimately, 29 breast cancer cases were identified, with 4 cases in the 35–44 years group and 25 cases in the 45–64 years group.
The distribution of malignant imaging findings across different BI-RADS classification.
| Method | Type | Malignant imaging findings | BI-RADS (n, %) | |||
|---|---|---|---|---|---|---|
| 1 | 2 | 3 | ≥4 | |||
| HHUS | Mass | Irregular Morphology | 0(0) | 43(7.5) | 279(48.5) | 251(44) |
| Indistinct margins | 0(0) | 4(1.9) | 34(16.1) | 171(81.0) | ||
| Not parallel (taller than wide) orientation | 0(0) | 1(4.5) | 5(22.7) | 16(72.7) | ||
| Calcification | grouped or linear | 0(0) | 0(0) | 0(0) | 4(100) | |
| Associated Features | Architectural distortion | 0(0) | 0(0) | 0(0) | 0(0) | |
| Skin thickening | 0(0) | 3(100) | 0(0) | 0(0) | ||
| Elastic imaging (hardness) | 0(0) | 0(0) | 0(0) | 0(0) | ||
| ABUS | Mass | Irregular Morphology | 0(0) | 42(7.9) | 246(46.4) | 242(45.7) |
| Indistinct margins | 0(0) | 2(1.1) | 40(21.4) | 145(77.5) | ||
| Not parallel (taller than wide) orientation | 0(0) | 0(0) | 2(11.1) | 16(88.9) | ||
| Calcification | grouped or linear | – | – | – | – | |
| Associated Features | Architectural distortion | 0(0) | 0(0) | 1(100) | 0(0) | |
| Skin thickening | 0(0) | 0(0) | 0(0) | 0(0) | ||
| Elastic imaging (hardness) | – | – | – | – | ||
| MG | Mass | Irregular Morphology | 0(0) | 0(0) | 7(22.6) | 24(77.4) |
| Indistinct margins | 0(0) | 0(0) | 42(59.2) | 29 (40.8) | ||
| High density | 0(0) | 0(0) | 9(45) | 11(55) | ||
| Calcification | grouped or linear | 0(0) | 1(14.3) | 0(0) | 6(85.7) | |
| Associated Features | Architectural distortion | 0(0) | 0(0) | 3(37.5) | 5(62.5) | |
| Skin thickening | 0(0) | 0(0) | 0(0) | 0(0) | ||
| Asymmetric density | 0(0) | 0(0) | 289(98.3) | 5(1.7) | ||
HHUS, Handheld ultrasound; ABUS, Automated breast ultrasound; CI, Confidence Interva; BI-RADS, the Breast Imaging-Reporting and Data System.
The agreement on BI-RADS classifications of category 4 or higher between HHUS and ABUS was 98.2%, with a Kappa coefficient of 0.726, indicative of a substantial agreement between the two imaging modalities. When stratified by age groups, the agreement for both 35–44 and 45–64 subgroups was found to be 98.2%. This suggests no significant variation in consistency between the two ultrasound techniques across different age groups (
The agreement of BI-RADS classification between HHUS and ABUS.
| Age group | HHUS | ABUS | Total | Agreement (%) | Kappa (95%CI) | |
|---|---|---|---|---|---|---|
| Positive | Negative | |||||
| 35–64 years | Positive | 225(2.5) | 84(0.9) | 309 | 98.2 | 0.726 |
| Negative | 78(0.9) | 8607(99.1) | 8685 | |||
| 35–44 years | Positive | 64(70.3) | 27(29.7) | 91 | 98.2 | 0.694 |
| Negative | 27(0.9) | 2918(99.1) | 2945 | |||
| 45–64 years | Positive | 161(73.9) | 57(26.1) | 218 | 98.2 | 0.739 |
| Negative | 51(0.9) | 5689(99.1) | 5740 | |||
Positive indicates BI-RADS category 4 or higher. Negative indicates BI-RADS category 3 or lower.
HHUS, Handheld ultrasound; ABUS, Automated breast ultrasound; CI, Confidence Interval.
In contrast, the agreement between ultrasound and MG within the 45–64 age group was comparatively lower, with agreement rates of 96.1% for HHUS and MG, and 96.3% for ABUS and MG, respectively. This discrepancy primarily stemmed from a higher proportion of cases where ultrasound flagged as positive while MG flagged as negative (77.3% for HHUS and 77.6% for ABUS), versus fewer instances where ultrasound flagged as negative while MG flagged as positive (22.7% for HHUS and 22.4% for ABUS) (
The agreement of BI-RADS classification between ultrasound and MG in 45–64 years group.
| Method | Result | MG | Total | Agreement (%) | Kappa (95%CI) | |
|---|---|---|---|---|---|---|
| Positive | Negative | |||||
| HHUS | Positive | 35(16.3) | 180(83.7) | 215 | 96.1 | 0.2145 |
| Negative | 53(0.9) | 5670(99.1) | 5723 | |||
| ABUS | Positive | 39(18.7) | 170(81.3) | 209 | 96.3 | 0.2469 |
| Negative | 49(0.9) | 5682(99.1) | 5731 | |||
Positive indicates BI-RADS category 4 or higher. Negative indicates BI-RADS category 3 or lower.
HHUS, Handheld ultrasound; ABUS, Automated breast ultrasound; MG, Mammography; CI, Confidence Interval.
The efficacy of different image-based BI-RADS classification systems in breast cancer screening was evaluated across age groups. In the 35–44 age group, both HHUS and ABUS independently detected all 4 breast cancer cases of, exhibiting comparable performance metrics in sensitivity(75% vs 75%), specificity(98.55% vs 98.58%), PPV(6.52% vs 6.67%), and NPV(99.97% vs 99.97%). Among the 45–64 years group, ABUS detected all 25 confirmed breast cancer cases, HHUS detected 23 cancer cases, and MG only identified 20 cancer cases. The sensitivities were 100.0% for ABUS, 92.0% for HHUS, and 80.0% for MG, with specificities ranging from 98.6% to 99.7% (
Performance metrics of the BI-RADS classification in cancer screening.
| Method | 35–44 years group | 45–64 years group | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Sensitivity | Specificity | PPV | NPV | AUC | Sensitivity | Specificity | PPV | NPV | AUC | |
| HHUS | 75.00 |
98.55 |
6.52 |
99.97 |
0.8678 | 92.00 |
98.56 |
21.70 |
99.96 |
0.9529 |
| ABUS* | 75.00 |
98.58 |
6.67 |
99.97 |
0.8679 | 100 |
98.66 |
24.51 |
100 |
0.9935 |
| MG | – | – | – | – | 80.00 |
99.65 |
50.00 |
99.91 |
0.8983 | |
| HHUS+MG* | – | – | – | – | 100 |
98.39 |
21.37 |
100 |
0.9920 | |
| ABUS+MG* | – | – | – | – | 100 |
98.55 |
23.15 |
100 |
0.9928 | |
HHUS, Handheld ultrasound; ABUS, Automated breast ultrasound; MG, Mammography; PPV, Positive predictive value; NPV, Negative Predictive value; AUC, The area under the receiver operating characteristic (ROC) curve; HHUS+MG HHUS used for initial screening, HHUS-negative women undergo MG screening, ABUS+MG ABUS is used for initial screening, and MG screening is used for ABUS-negative women.
*Compared with mammography alone, significant differences were observed for AUC in the 45–64 years group.
This study found that almost all malignant signs identified by radiologists at primary healthcare hospitals were diagnosed as BI-RADS category 3 or higher, which alert clinical attention. Signs with high malignancy indications, such as not parallel (taller than wide) orientation and grouped calcifications, were categorized as higher, namely BI-RADS category 4 or higher, to prompt timely biopsies in clinical practice. Ultrasound is sensitive in detecting mass-related malignant signs, while MG is sensitive in detecting malignant calcifications and architectural distortion. The efficacy of BI-RADS classification used in breast cancer screening is promising in primary healthcare settings.
In this study, it was observed that irregularly shaped masses were similarly categorized as BI-RADS category 3 and category 4 or higher by HHUS and ABUS. However, in MG, these masses tended to be assigned higher BI-RADS levels. Considering the significantly lower detection of mass lesions by MG compared to ultrasound, it can be hypothesized that this difference may be related to the lower sensitivity of MG in detecting isodense lesions. Therefore, irregularly shaped masses visualized on MG often draw high attention from radiologists (
In the context of breast cancer screening, calcifications are pivotal imaging findings, especially in MG examinations. Coarse or popcorn-like calcifications are often associated with benign lesions, while small and grouped, branching, or linear calcifications may indicate malignancy (
The study demonstrated a high diagnostic consistency between HHUS and ABUS, indicating that ABUS can be effectively utilized for BI-RADS classification like HHUS. However, when comparing HHUS and ABUS to MG, the consistency was lower with 96.1% and 96.3% respectively. Discrepancies predominantly manifested as a higher proportion of cases (77.3% for HHUS and 77.6% for ABUS) where both HHUS and ABUS detected abnormalities while MG did not, as opposed to fewer instances (22.7% for HHUS and 22.4% for ABUS) where ultrasound flagged as negative while MG flagged as positive. The differences may be due to the varying abilities of ultrasound and MG to detect different types of malignancy-related pathological features observed in this study. Ultrasound may be more effective in identifying mass-related signs, while MG may be more excellent at detecting calcifications and architectural distortions. The reason why MG detects fewer mass-related signs compared to ultrasound may be related to the high-density glandular tissue that can obscure isodense lesions, while ultrasound is not affected by such tissue density (
This study also evaluated the performance of BI-RADS classification system in breast cancer screening, highlighting notable differences across age groups and screening methods. In the screening program for individuals aged 35-44, both HHUS and ABUS showed similar diagnostic performance, with sensitivities of 75% and specificities of 98.55% and 98.58%, respectively. For individuals aged 45-64, ABUS showed slightly higher sensitivity (100% vs 92%) and specificity (98.66% vs 98.56%). This aligns with a similar finding in a previous study that ABUS has statistically significant higher diagnostic accuracy than HHUS in detecting breast cancer (
Nevertheless, this study has several limitations that should be considered when interpreting its results. Firstly, our study was conducted at a single center, which may limit the breadth of our findings. The practices of BI-RADS in primary healthcare in our study may not be fully representative of those found in other settings. Secondly, while our study evaluated the implementation of the BI-RADS classification system in primary healthcare hospitals, it did not involve a quality control assessment by highly experienced radiologists to directly compare and more accurately evaluate the standardization and compliance of BI-RADS application at these facilities.
Radiologists at primary hospitals can effectively adhere to BI-RADS guidelines to provide clinical indications of malignant risks. The differences in BI-RADS classification diagnoses between ultrasound and MG reflect the characteristics of each imaging technique. Based on the BI-RADS findings, HHUS, ABUS, and MG have good efficacy in breast cancer screening. In conclusion, the application of BI-RADS is acceptable in primary healthcare hospitals.
The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.
The studies involving humans were approved by ethic committee of Henan Cancer Hospital (Approval Number: 19/109-1893). The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation in this study was provided by the participants’ legal guardians/next of kin.
XL: Methodology, Writing – original draft, Project administration, Writing – review & editing, Conceptualization. HW: Methodology, Writing – review & editing, Data curation, Formal analysis, Funding acquisition. H-FX: Data curation, Formal analysis, Methodology, Funding acquisition, Writing – review & editing. S-KZ: Conceptualization, Formal analysis, Resources, Writing – review & editing, Methodology, Project administration, Data curation. B-JZ: Formal analysis, Writing – review & editing, Methodology, Validation. H-LL: Writing – review & editing, Supervision, Conceptualization, Project administration.
The author(s) declare financial support was received for the research and/or publication of this article. This project was supported by the Henan Key Project of Science and Technology (Grant number: 202102310107), the Henan Provincial Natural Science Foundation (Grant number: 242300420416), the Open Project of the Key Laboratory of Cancer Invasion and Metastasis, Ministry of Education (grant no. 2024KFKT009), and the Henan Province Key R&D and Promotion Project (Grant number: 232102310242).
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The author(s) declare that Generative AI was used in the creation of this manuscript. AI only used for academic translation of manuscripts and grammar correction.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.