Mammographic classification of interval breast cancers and artificial intelligence performance - Article

Q: How were interval breast cancers identified in this study?

Interval breast cancers were diagnosed less than 12 months after a preceding negative mammogram, using screening mammograms acquired between 2010 and 2019.

Q: What categories did radiologists use for interval breast cancer classification?

Radiologists classified interval breast cancers as missed (reading error, minimal signs, actionable/nonactionable), true interval, occult, or missed due to technical error.

Q: How did the AI tool perform and what did “flagged” mean?

The deep-learning AI assigned risk scores from 1 to 10, with scores of 8 or higher considered “flagged,” and it localized mammographically visible cancer types more accurately than nonvisible types.

Sep 23, 2026

Uploaded by McGucket

Study evaluates mammographic classification of interval breast cancers and the performance of a deep-learning artificial intelligence tool in the United States, emphasizing a 12-month screening interval. Interval breast cancers were defined as diagnoses occurring less than 12 months after a negative prior mammogram, using 2010–2019 digital mammography and digital breast tomosynthesis. Radiologists retrospectively categorized cancers into missed (reading error, technical error), minimal signs (actionable, nonactionable), true interval, and occult types. Results assess AI risk scoring (1–10), flagged exams (≥8), localization accuracy, and associations with patient and tumor characteristics.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Abstract

Background

Methods

Results

Conclusion

Introduction

JNCI: Journal of the National Cancer Institute, 2025, 117(8), 1627–1638 [https://doi.org/10.1093/jnci/djaf103](https://doi.org/10.1093/jnci/djaf103) Advance Access Publication Date: April 18, 2025 Article Mammographic classification of interval breast cancersand artificial intelligence performance Tiffany T. Yu , MD􀀃 ,1 ,2 , Anne C. Hoyt , MD1, 2, Melissa M. Joines, MD1, 2, Cheryce P. Fischer, MD1, 2, Nazanin Yaghmai, MD1,2, James S. Chalfant , MD1,2, Lucy Chow , MD1, 2, Shabnam Mortazavi, MD1, 2, Christopher D. Sears , BS1,2, James Sayre , PhD1, 2, Joann G. Elmore , MD, MPH1, 3, William Hsu , PhD1, 2, Hannah S. Milch, MD1,2 1David Geffen School of Medicine, University of California Los Angeles, Los Angeles, CA, United States 2Department of Radiological Sciences, University of California Los Angeles Health Sciences, Los Angeles, CA, United States 3Department of Medicine, University of California Los Angeles Health Sciences, Los Angeles, CA, United States 􀀃 Corresponding author: Tiffany T. Yu, MD, Department of Radiological Sciences, UCLA Santa Monica Hospital, Radiology Administration, 1250 16th St, Ste 2340, Santa Monica, CA 90404, United States ([ttyu@mednet.ucla.edu](ttyu@mednet.ucla.edu)) . Abstract Background: European studies suggest that artificial intelligence (AI) can reduce interval breast cancers. Research on interval breast cancer classification and AI’s effectiveness in the United States, however, particularly using digital breast tomosynthesis and annual screening, is limited. We aimed to mammographically classify interval breast cancers and assess AI performance using a 12-month screening interval. Methods: From digital mammography and digital breast tomosynthesis screening mammograms acquired between 2010 and 2019 at a US tertiary-care academic center, we identified interval breast cancers diagnosed less than 12 months after a negative mammogram. At least 3 breast radiologists retrospectively classified interval breast cancers as missed—reading error, minimal signs—actionable, minimal signs—nonactionable, true interval, occult, or missed—technical error. A deep-learning AI tool assigned risk scores ranging from 1 to 10 to the negative index screening mammograms, with scores of 8 or higher considered “flagged.” Statistical analysis evaluated associations among interval breast cancer types and AI exam scores, AI markings, and patient and tumor characteristics. Results: From 184 935 screening mammograms (65% digital mammography, 35% digital breast tomosynthesis), we identified 148 interval breast cancers in 148 women (mean [SD] age ¼ 61 [12] years) . Of these, 26% were minimal signs—actionable, 24% were occult, 22% were minimal signs—nonactionable, 17% were missed—reading error, 6% were true interval, and 5% were missed—technical error (P < .001). AI scored 131 mammograms (17 errors excluded); it most frequently flagged exams with missed—reading error (90%), minimal signs—actionable (89%), and minimal signs—nonactionable (72%) (P ¼ .02) . AI localized mammographically visible types more accurately (35%-68%) than nonvisible types (0%-50%; P ¼ .02) . Conclusion: AI more frequently flagged and accurately localized interval breast cancer types that were mammographically visible at screening (missed or minimal signs) compared with true interval or occult cancers. Introduction Interval breast cancers are diagnosed between scheduled screening mammograms, with the preceding mammogram interpreted as negative for malignancy.1 Screening intervals vary from 1 to 3 years, depending on program guidelines. Interval breast cancers are associated with a poorer prognosis than screen-detected cancers.2 , 3 Thus, reducing interval breast cancer rates is a strong surrogate for assessing screening program effectiveness. Specifically, it is important to understand whether interval breast cancers were “missed” by the radiologist or just not visible at the time of the preceding screening mammogram. Although

FAQs

How were interval breast cancers identified in this study?

Interval breast cancers were diagnosed less than 12 months after a preceding negative mammogram, using screening mammograms acquired between 2010 and 2019.

What categories did radiologists use for interval breast cancer classification?

Radiologists classified interval breast cancers as missed (reading error, minimal signs, actionable/nonactionable), true interval, occult, or missed due to technical error.

How did the AI tool perform and what did “flagged” mean?

The deep-learning AI assigned risk scores from 1 to 10, with scores of 8 or higher considered “flagged,” and it localized mammographically visible cancer types more accurately than nonvisible types.

Sep 23, 2026

Uploaded by McGucket