Article Tools
Selective artificial intelligence arbitration by BI-RADS category: an algorithm to reduce overdiagnosis in mammographic breast cancer screening
Keywords:
breast cancer, screening mammography, artificial intelligence, BI-RADS, false-positive results, clinical decision support, UzbekistanAbstract
Background. Mammography screening in Uzbekistan is moving to hybrid reading by a radiologist and an artificial intelligence (AI) system. How the two should be combined in a setting where radiologist sensitivity is already high but false-positive recalls are frequent has not been studied.
Aim. To develop and evaluate an algorithm in which the AI system reviews the radiologist’s assessment only in the Breast Imaging Reporting and Data System (BI-RADS) categories where that assessment is least reliable.
Materials and methods. Retrospective diagnostic accuracy study of 1,759 digital mammography examinations from Tashkent, Nukus and Samarkand with a verified outcome (316 breast cancers, 18.0%). Radiologists’ BI-RADS assessments and the continuous risk score of a national AI system were obtained independently and compared with histopathology or 6–24-month follow-up. The algorithm was derived from the structure of radiologist errors and reader–AI disagreement: BI-RADS 4B–5 assessments are kept, the AI arbitrates BI-RADS 0 and 4A (threshold 0.30) and acts as a safety net for BI-RADS 1–3 (threshold 0.65). Paired comparisons used McNemar’s test.
Results. Of 302 false-positive radiologist assessments, 277 (91.7%) fell into BI-RADS 0, 4A and 4B. All 6 cancers missed by radiologists scored above the AI operating threshold, and no cancer occurred among 982 examinations rated negative by both. The algorithm removed 70 unjustified positive assessments; specificity rose from 79.1% to 83.9% and positive predictive value from 50.7% to 57.3%, while sensitivity was unchanged (98.1% vs 98.4%; p = 1.000). The error rate fell from 17.5% to 13.5% (difference 4.0 percentage points, 95% CI 2.8–5.3; p < 0.001), and the reduction was seen in every density, age and regional subgroup.
Conclusion. Using AI only where radiologist uncertainty is highest reduces overdiagnosis without loss of cancer detection. The algorithm needs prospective validation in a population screening cohort.
References
1. Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71(3):209–249. doi:10.3322/caac.21660
2. Tillyashaykhov MN, Djanklich SM, Ibragimov SN, Imamov OA. Analysis of cancer incidence structure in the Republic of Uzbekistan. Oncology and Radiology of Kazakhstan. 2021;3(61):4–8. doi:10.52532/2663-4864-2021-3-61-4-8
3. World Health Organization. Global Breast Cancer Initiative implementation framework: assessing, strengthening and scaling-up of services for the early detection and management of breast cancer. Geneva: WHO; 2023.
4. Resolution of the President of the Republic of Uzbekistan No. PP-402 of 22 November 2024 on measures for the early detection of breast and cervical cancer.
5. Elmore JG, Wells CK, Lee CH, Howard DH, Feinstein AR. Variability in radiologists' interpretations of mammograms. N Engl J Med. 1994;331(22):1493–1499. doi:10.1056/NEJM199412013312206
6. Kerlikowske K, Grady D, Barclay J, et al. Variability and accuracy in mammographic interpretation using the American College of Radiology Breast Imaging Reporting and Data System. J Natl Cancer Inst. 1998;90(23):1801–1809. doi:10.1093/jnci/90.23.1801
7. Lehman CD, Arao RF, Sprague BL, et al. National performance benchmarks for modern screening digital mammography: update from the Breast Cancer Surveillance Consortium. Radiology. 2017;283(1):49–58. doi:10.1148/radiol.2016161174
8. Perry N, Broeders M, de Wolf C, et al. European guidelines for quality assurance in breast cancer screening and diagnosis. Fourth edition – summary document. Ann Oncol. 2008;19(4):614–622. doi:10.1093/annonc/mdm481
9. McKinney SM, Sieniek M, Godbole V, et al. International evaluation of an AI system for breast cancer screening. Nature. 2020;577(7788):89–94. doi:10.1038/s41586-019-1799-6
10. Salim M, Wåhlin E, Dembrower K, et al. External evaluation of 3 commercial artificial intelligence algorithms for independent assessment of screening mammograms. JAMA Oncol. 2020;6(10):1581–1588. doi:10.1001/jamaoncol.2020.3321
11. Yoon JH, Strand F, Baltzer PAT, et al. Standalone AI for breast cancer detection at screening digital mammography and digital breast tomosynthesis: a systematic review and meta-analysis. Radiology. 2023;307(5):e222639. doi:10.1148/radiol.222639
12. Lång K, Josefsson V, Larsson AM, et al. Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (MASAI): a clinical safety analysis of a randomised, controlled, non-inferiority, single-blinded, screening accuracy study. Lancet Oncol. 2023;24(8):936–944. doi:10.1016/S1470-2045(23)00298-X
13. Dembrower K, Crippa A, Colón E, Eklund M, Strand F. Artificial intelligence for breast cancer detection in screening mammography in Sweden: a prospective, population-based, paired-reader, non-inferiority study. Lancet Digit Health. 2023;5(10):e703–e711. doi:10.1016/S2589-7500(23)00153-X
14. Eisemann N, Bunk S, Mukama T, et al. Nationwide real-world implementation of AI for cancer detection in population-based mammography screening. Nat Med. 2025;31(3):917–924. doi:10.1038/s41591-024-03408-6
15. Lauritzen AD, Rodríguez-Ruiz A, von Euler-Chelpin MC, et al. An artificial intelligence-based mammography screening protocol for breast cancer: outcome and radiologist workload. Radiology. 2022;304(1):41–49. doi:10.1148/radiol.210948
16. Rodriguez-Ruiz A, Lång K, Gubern-Merida A, et al. Can we reduce the workload of mammographic screening by automatic identification of normal exams with artificial intelligence? A feasibility study. Eur Radiol. 2019;29(9):4825–4832. doi:10.1007/s00330-019-06186-9
17. Tokhtamuratov U, Adilkhodjaev A, Ibragimov I, Kurbanova S, Kasparova G, Seytmuratova A, Ososkov A. Deep learning-based breast cancer detection on mammography in Uzbekistan: a multicentre retrospective study. Klinichna Onkologiya (Clinical Oncology). 2026;16(4):1–8. doi:10.32471/clinicaloncology.2663-466X.36794
18. American College of Radiology. ACR BI-RADS Atlas: Breast Imaging Reporting and Data System. 5th ed. Reston, VA: American College of Radiology; 2013.
19. Leibig C, Brehmer M, Bunk S, et al. Combining the strengths of radiologists and AI for breast cancer screening: a retrospective analysis. Lancet Digit Health. 2022;4(7):e507–e519. doi:10.1016/S2589-7500(22)00070-X
20. Kwon MR, Chang Y, Ham SY, et al. Screening mammography performance according to breast density: a comparison between radiologists versus standalone intelligence detection. Breast Cancer Res. 2024;26(1):68. doi:10.1186/s13058-024-01821-w
21. Kim HE, Kim HH, Han BK, et al. Changes in cancer detection and false-positive recall in mammography using artificial intelligence: a retrospective, multireader study. Lancet Digit Health. 2020;2(3):e138–e148. doi:10.1016/S2589-7500(20)30003-0
22. Houssami N, Marinovich ML. AI for mammography screening: enter evidence from prospective trials. Lancet Digit Health. 2023;5(10):e641–e642. doi:10.1016/S2589-7500(23)00176-0
23. Al-Bazzaz H, Janicijevic M, Strand F. Reader bias in breast cancer screening related to cancer prevalence and artificial intelligence decision support – a reader study. Eur Radiol. 2024;34(8):5415–5424. doi:10.1007/s00330-023-10514-5
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Askar Adilkhodjaev, Umid Tokhtamuratov (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.