Comparison of mammography breast density assessment by radiologists and Quantra artificial intelligence in women with and without breast implants: intra and interobserver agreement
Introduction: Visual methods for assessing mammographic breast density (MBD) have variable reproducibility in women with dense breasts and/or implants. Assessment of MBD by artificial intelligence (AI) and radiologists (human readers, HRs) has been reported. This study compared the intraobserver agr...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Permanyer
2025-04-01
|
| Series: | Journal of the Mexican Federation of Radiology and Imaging |
| Online Access: | https://www.jmexfri.com/frame_eng.php?id=140 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Introduction: Visual methods for assessing mammographic breast density (MBD) have variable reproducibility in women with dense breasts and/or implants. Assessment of MBD by artificial intelligence (AI) and radiologists (human readers, HRs) has been reported. This study compared the intraobserver agreement of HRs and the interobserver agreement of Quantra AI software and HRs in MBD assessment in women with and without breast implants. Material and methods: HRs and AI Quantra 2.2.2 assessed MBD in 2D mammograms and tomosynthesis of women over 35 years of age in two phases. Four (a, b, c, d) and two (a+b, non-dense and c+d, dense) MBD categories of the Breast Imaging Reporting and Data System (BI-RADS) were used. Intra and interobserver agreement was assessed with light and Cohen’s kappa. Results: 4-HRs evaluated 678 mammograms in each phase: 548 without implants and 130 with implants. 4-HRs intraobserver agreement was substantial to almost perfect in women without implants in the two MBD categories. For women with implants, intraobserver agreement was moderate to substantial in the four and two MBD categories. In women without implants, interobserver agreement between AI Quantra and 4-HRs was moderate to substantial for the two MBD categories, while in women with implants, interobserver agreement was slight to fair for the four and two MBD categories. Conclusion: Intraobserver agreement for the 4-HRs was better in women without implants in the two MBD categories and acceptable in women with implants. The interobserver agreement between AI Quantra and 4-HRs was acceptable for the two MBD categories in women without implants, while in women with implants, the interobserver agreement between AI Quantra and 4-HRs was unacceptable for the four and two MBD categories. This is the first report on the comparison of AI Quantra and HRs in the assessment of MBD in women with and without breast implants.
|
|---|---|
| ISSN: | 2938-1215 2696-8444 |