AI can be used alone for low-malignancy mammograms

Kate Madden Yee, Senior Editor, AuntMinnie.com. Headshot
Kate Madden Yee
Nov 4, 2019
2017 11 07 18 37 9362 Roadies Ribbon 400

Sunday, December 1 | 11:15 a.m.-11:25 a.m. | SSA01-04 | Room S406A

An artificial intelligence (AI) algorithm can be used for standalone interpretation of mammograms that have a low probability of being malignant, researchers from the University of Southern California will report in this presentation.

Studies have shown that the accuracy of AI-based algorithms for 2D mammography can match or exceed that of the average radiologist, presenter Dr. Alyssa Watanabe and colleagues noted. The group conducted a study to test this conclusion by evaluating the accuracy of an AI-based algorithm set to a 99% sensitivity threshold on a set of 1,255 screening mammograms.

The exams consisted of both benign and malignant cases. Out of the total pool of cases, the AI algorithm classified 40% as not suspicious; no cancers were found in the cases when the team compared them with biopsy and/or long-term follow-up data. In addition, the algorithm categorized 99% of biopsy-proven cancers as suspicious.

"Using a high sensitivity threshold, it is possible that AI-based software could potentially be used as a standalone to eliminate very low probability for malignancy cases from the radiologist worklist," Watanabe and colleagues concluded.

This paper received a Roadie 2019 award for the most popular abstract by page views in this Road to RSNA section.

Latest in Breast
(A) Free-response receiver operating characteristic (FROC) curve for DBT-DINO and DINOv2 (Meta AI) on the detection task. DINOv2 achieved higher sensitivity than DBT-DINO across most false-positive thresholds. (B) Detection rates of DBT-DINO and DINOv2 for lesions that were ultimately found to be benign (n = 70) and lesions that were found to be cancerous (n = 66). Error bars represent 95% CIs estimated via patient-level bootstrapping with 1000 repetitions. Both models showed similar detection rates for cancerous lesions, whereas DINOv2 detected more benign lesions than DBT-DINO. (C) Example bounding box predictions for DBT-DINO and DINOv2 on the detection test set. Each image is a single two-dimensional section from a three-dimensional digital breast tomosynthesis volume; the first column shows a mediolateral oblique view, and the second and third columns show craniocaudal views. Both models struggled to detect very small lesions, such as the 3-mm-diameter lesion shown in the first column. FP = false positive.
DBT-based foundational model classifies breast density
August 4, 2026
Rad Site
RadSite publishes clinical coverage guideline for breast CBCT imaging
July 30, 2026
Images in a 70-year-old asymptomatic woman with a prior excisional biopsy in the left breast yielding lobular carcinoma in situ 3 years earlier. Screening contrast-enhanced mammography (CEM) was performed, which was considered a prevalence screen as the woman had no prior breast MRI in the past 3 years and no prior CEM. (A) Low-energy (26–30 kVp) mediolateral oblique (MLO) and (B) craniocaudal (CC) views show surgical clips but no abnormality. (C) Recombined MLO and (D) CC images show an enhancing mass in the medial left breast (arrows). Biopsy revealed invasive lobular carcinoma. Size was 0.2 cm; tumor grade II/III; negative lymph nodes; and estrogen receptor–positive, progesterone receptor–positive, and human epidermal growth factor receptor 2–negative.
CEM finds breast cancer at baseline, incidence screening
July 28, 2026
Minnie Cast Thumbnail (15)
Women's Imaging MinnieCast, Episode 13: Partial breast irradiation in the spotlight
July 17, 2026
Related Stories
Rsna
Market research reports
RSNA
2016 09 14 15 25 23 643 Dollar Sign 400
Breast
Do payouts influence mammography guidelines?
2017 11 02 17 42 1107 Road To Rsna 400
RSNA
Road to RSNA 2019: Women's Imaging Preview
2018 07 25 21 27 0157 Patient Woman Doctor 400
Breast
Same-day breast biopsies reduce healthcare disparities
More in Breast
DBT-based foundational model classifies breast density
By Amerigo Allegretto
DBT-DINO showed strong performance in classifying breast density from DBT images.
August 4, 2026
(A) Free-response receiver operating characteristic (FROC) curve for DBT-DINO and DINOv2 (Meta AI) on the detection task. DINOv2 achieved higher sensitivity than DBT-DINO across most false-positive thresholds. (B) Detection rates of DBT-DINO and DINOv2 for lesions that were ultimately found to be benign (n = 70) and lesions that were found to be cancerous (n = 66). Error bars represent 95% CIs estimated via patient-level bootstrapping with 1000 repetitions. Both models showed similar detection rates for cancerous lesions, whereas DINOv2 detected more benign lesions than DBT-DINO. (C) Example bounding box predictions for DBT-DINO and DINOv2 on the detection test set. Each image is a single two-dimensional section from a three-dimensional digital breast tomosynthesis volume; the first column shows a mediolateral oblique view, and the second and third columns show craniocaudal views. Both models struggled to detect very small lesions, such as the 3-mm-diameter lesion shown in the first column. FP = false positive.
RadSite publishes clinical coverage guideline for breast CBCT imaging
By AuntMinnie.com staff writers
The guide outlines how breast CBCT should be used when conventional imaging approaches are inconclusive or inaccessible.
July 30, 2026
Rad Site
CEM finds breast cancer at baseline, incidence screening
By Amerigo Allegretto
Incidence screening with CEM maintains high performance and may offer more benefits over baseline CEM screening.
July 28, 2026
Images in a 70-year-old asymptomatic woman with a prior excisional biopsy in the left breast yielding lobular carcinoma in situ 3 years earlier. Screening contrast-enhanced mammography (CEM) was performed, which was considered a prevalence screen as the woman had no prior breast MRI in the past 3 years and no prior CEM. (A) Low-energy (26–30 kVp) mediolateral oblique (MLO) and (B) craniocaudal (CC) views show surgical clips but no abnormality. (C) Recombined MLO and (D) CC images show an enhancing mass in the medial left breast (arrows). Biopsy revealed invasive lobular carcinoma. Size was 0.2 cm; tumor grade II/III; negative lymph nodes; and estrogen receptor–positive, progesterone receptor–positive, and human epidermal growth factor receptor 2–negative.
Women's Imaging MinnieCast, Episode 13: Partial breast irradiation in the spotlight
By Amerigo Allegretto
Two committee members for ASTRO's PBI guidelines share insights about this breast cancer treatment option.
July 17, 2026
Minnie Cast Thumbnail (15)
Incorrect AI suggestions influence reader performance on mammography
By Amerigo Allegretto
The biggest impact on reader performance stemmed from false-negative AI suggestions.
July 14, 2026
A normal mammogram confirmed by three-year radiologic follow-up illustrates reader-marked regions of interest (ROIs) during (A) unaided (round 1) and (B) artificial intelligence (AI)–assisted (round 2) reading. Each colored dot represents an ROI for recall by a human reader. Readers could mark more than one ROI per case, represented by multiple dots of the same color. During AI-assisted reading, the AI system displayed three visible prompts: two with suspicion of malignancy scores of 35% (left mediolateral oblique [L MLO] and craniocaudal [L CC]) and one with a suspicion of malignancy score of 10% (right craniocaudal [R CC]), shown as polygonal overlays. Without AI, six of 10 readers (60%) marked a false-positive ROI. With AI assistance, this fell to two of 10 (20%). R MLO = right mediolateral oblique.
Breast cancer screening higher among civically engaged women
By Amerigo Allegretto
Women who voted, attended a public meeting, or volunteered are more likely to get their breast cancer screening.
July 14, 2026
Mammogram Calendar 400
CEM on par with MRI for breast cancer diagnosis, local staging
By Amerigo Allegretto
CEM alone in a prospective study achieved perfect sensitivity and negative predictive value in diagnosing breast cancer.
July 10, 2026
MRI false positive for additional disease, (top to bottom): A. Index lesion—grade 2 invasive ductal carcinoma; B. MRI lesion two, upper inner quadrant: Stromal sclerosis and columnar cell change; C. MRI lesion three, upper outer quadrant: No ultrasound correlate, no malignancy on follow-up; D. CEM marked background parenchymal enhancement with a unifocal retroareolar lesion demonstrating wash-out on delayed MLO. Images are republished under a Creative Commons license (CC BY 4.0).
Siemens Healthineers gets FDA nod for CEM on Mammomat B.brilliant
By AuntMinnie.com staff writers
Siemens Healthineers has received FDA clearance for CEM and CEM biopsy capabilities for its Mammomat B.brilliant system.
June 30, 2026
Mammomat B.brilliant breast imaging system
Fujifilm, IAEA sign agreement to expand cancer imaging
By AuntMinnie.com staff writers
The two organizations will collaborate on awareness campaigns, education and training programs for healthcare professionals, and the provision of Fujifilm medical equipment.
June 29, 2026
Acquisitions
DeepHealth gets FDA nod for two mammography AI tools
By AuntMinnie.com staff writers
DeepHealth has received FDA clearance for two parts of the company's Breast Suite AI platform: BAC Assessment and Mammo Dx.
June 25, 2026
Deephealth White On Purple
QT Imaging reports low variability for breast scanner biomarkers
By AuntMinnie.com staff writers
Study analyzes Breast Acoustic CT speed-of-sound (SoS) measurements.
June 24, 2026
Breast Update 2 T76 Vuzvgq
Komen awards $15.4 million in breast cancer research grants
By AuntMinnie.com staff writers
Susan G. Komen has awarded over $15 million in grant funding toward breast cancer research.
June 23, 2026
Research Money 400
Page 1 of 584
Next Page