Landmark Nature Health Study Demonstrates the Effectiveness of DeepHealth’s Novel AI-Powered Breast Cancer Detection Workflow

Rhea-AI Summary

RadNet (NASDAQ: RDNT) and DeepHealth reported results from the ASSURE study, the largest U.S. real-world analysis of AI-driven breast cancer screening, published in Nature Health on Nov 17, 2025. The study evaluated >579,000 women across 109 community imaging sites and found the AI-powered workflow in RadNet’s Enhanced Breast Cancer Detection (EBCD) program produced a 21.6% increase in cancer detection rate vs state-of-the-art 3D mammography, a 15% increase in positive predictive value, and maintained recall rates within ACR guidelines. Benefits were consistent across populations, including >150,000 Black women and a 22.7% boost for women with dense breasts.

Positive

• Sample size of >579,000 women across 109 sites
• Cancer detection +21.6% vs 3D mammography
• PPV +15% while keeping recall within ACR guidelines
• Dense-breast detection +22.7% vs 3D mammography
• Included >150,000 Black women, showing consistent benefits

Negative

• None.

Insights

Clinical Imaging Strategist

Large real‑world study shows DeepHealth’s AI workflow raised cancer detection while keeping recall rates within guideline limits.

The analysis of over 579,000 women across 109 community sites, published on Nov. 17, 2025, found the AI‑supported workflow increased cancer detection rate by 21.6% versus 3D mammography and raised positive predictive value by 15%. The workflow combined DeepHealth’s FDA‑cleared CADe/x with an AI‑supported Safeguard Review that can trigger expert re‑review, and RadNet reports this approach has run nationwide since 2023.

This result implies an operational mechanism where AI flags higher‑suspicion exams and routes them for additional human review, yielding more true positives without breaching American College of Radiology recall guidelines. The study reports consistent benefits across demographic groups, including more than 150,000 Black women and a 22.7% detection boost for women with dense breasts, while noting Black women face 40% higher breast cancer mortality in the U.S.

Key dependencies and risks include sustained real‑world implementation fidelity (AI thresholds, review workflow adherence) and continued maintenance of recall and positive predictive value within guideline ranges. The nationwide rollout since 2023 and publication in Nature Health support real‑world applicability, but ongoing monitoring of operational metrics is necessary to confirm persistence of these effects.

Watch for continued outcome reporting from the EBCD™ program and any site‑level performance summaries at upcoming meetings such as RSNA 2025; near‑term signals (months) would include site adoption rates and aggregated detection/recall metrics, while longer‑term signals (1–2 years) would show whether increased detection translates into stage‑shift or survival benefits.

Largest real-world analysis of AI-driven breast cancer screening in U.S. history¹ demonstrates increased cancer detection rate with consistent benefits across patient populations

LOS ANGELES and SOMERVILLE, Mass., Nov. 17, 2025 (GLOBE NEWSWIRE) -- RadNet, Inc. (NASDAQ: RDNT), the nation's largest provider of outpatient diagnostic imaging services, and its wholly owned subsidiary, DeepHealth, a global leader in AI-powered health informatics, today announced results from the largest real-world analysis of AI-driven breast cancer screening ever conducted in the United States.¹ Published in Nature Health, the findings support the clinical effectiveness and benefit of DeepHealth’s AI technology to deliver equitable results across several racial, ethnic and breast density patient groups.

The AI-Supported Safeguard Review Evaluation (ASSURE) study examined the AI-powered workflow that is at the heart of RadNet’s Enhanced Breast Cancer Detection™ (EBCD™) program. This study included mammograms from over 579,000 women across 109 community-based imaging sites in California, Delaware, Maryland and New York. The research compared state-of-the-art 3D mammography screening to a novel AI-driven protocol that combines DeepHealth’s FDA-cleared computer-aided detection and diagnosis (CADe/x) software with an AI-supported Safeguard Review workflow, which can trigger a second breast imaging expert review of high-suspicion cases—a workflow that RadNet now offers as EBCD™.

“Beyond the remarkable results, what sets this research apart is its scale, diversity and real-world applicability,” said Dr. Howard Berger, President and Chief Executive Officer of RadNet. “There has never been a similar study of this size in the United States, much less one with such a diverse patient population, that examines the patient impact and efficacy of AI-assisted breast cancer screening.”

The ASSURE study demonstrated that the AI-powered workflow led to a 21.6% increase in cancer detection rate compared to state-of-the-art 3D mammography screening, while maintaining recall rates within American College of Radiology guidelines² and increasing positive predictive value by 15%. This workflow is enabled by the applications that make up DeepHealth’s Breast Suite offering. Together, they deliver these benefits across patient populations, including the more than 150,000 Black women enrolled. Black women face 40% higher breast cancer mortality in the United States.³ Furthermore, the ASSURE study showed that the workflow underlying RadNet’s EBCD™ program delivered a 22.7% boost in cancer detection rate compared to 3D mammography screening for women with dense breasts, who experience both increased cancer risk and diagnostic challenges.⁴

“Unlike many academically focused studies, these screenings took place at community imaging centers, where most women get their mammograms,” said Dr. Gregory Sorensen, co-author of the ASSURE study and Chief Science Officer at RadNet. “To avoid potential selection bias, the AI-enabled workflow was provided to all patients at no additional charge during the study period. These real-world findings demonstrate how AI can improve access to specialist-level care for women, no matter where they live. When breast cancer is found early, women have far more options for care.”

Launched nationwide at RadNet-affiliated centers in 2023, EBCD™ runs on the AI that powers the applications within DeepHealth’s Breast Suite, helping detect lesions that are suspicious of being cancer, including those that are considered more difficult to find.⁶ Learn more about EBCD™ at myebcdmammo.com and discover Breast Suite at the DeepHealth booth (#1329 South Hall) at the Radiological Society of North America 2025 Annual Meeting (RSNA 2025).

About RadNet, Inc.
RadNet, Inc. is a leading national provider of freestanding, fixed-site diagnostic imaging services in the United States based on the number of locations and annual imaging revenue. RadNet has a network of 407 owned and/or operated outpatient imaging centers. RadNet’s markets include Arizona, California, Delaware, Florida, Maryland, New Jersey, New York and Texas. In addition, RadNet provides radiology information technology and artificial intelligence solutions marketed under the DeepHealth brand, teleradiology professional services and other related products and services to customers in the diagnostic imaging industry. Together with contracted radiologists, and inclusive of full-time and per diem employees and technologists, RadNet has about 11,000 team members. For more information, visit radnet.com.

About DeepHealth
DeepHealth is a wholly owned subsidiary of RadNet, Inc. (NASDAQ: RDNT) and serves as the umbrella brand for all companies within RadNet’s Digital Health segment. DeepHealth provides AI-powered health informatics with the aim of empowering breakthroughs in care through imaging. Building on the strengths of the companies it has integrated and is rebranding (e.g., CIMAR UK cloud-native healthcare image management solutions, eRAD Radiology Information and Image Management Systems and Picture Archiving and Communication System, Aidence lung AI, DeepHealth, Kheiron and iCAD breast AI, Quantib prostate and brain AI, and See-Mode thyroid and breast AI), DeepHealth leverages advanced AI for operational efficiency and improved clinical outcomes in brain, breast, chest, prostate, and thyroid health. At the heart of DeepHealth’s portfolio is a cloud-native operating system – DeepHealth OS – that unifies data across the clinical and operational workflow and personalizes AI-powered workspaces for everyone in the radiology continuum. Thousands of imaging centers and radiology departments around the world use DeepHealth solutions to enable earlier, more reliable, and more efficient disease detection, including in large-scale cancer screening programs. DeepHealth’s human-centered, intuitive technology aims to push the boundaries of what’s possible in healthcare. https://deephealth.com/

Forward-Looking Statements

This communication contains certain “forward-looking statements” within the meaning of the safe harbor provisions of the U.S. Private Securities Litigation Reform Act of 1995, Section 27A of the Securities Act of 1933, as amended, and Section 21E of the Securities Exchange Act of 1934, as amended. Forward-looking statements can be identified by words such as: “anticipate,” “believe,” “could,” “estimate,” “expect,” “forecast,” “intend,” “may,” “outlook,” “plan,” “potential,” “possible,” “predict,” “project,” “seek,” “should,” “target,” “will” or “would,” the negative of these words, and similar references to future periods. Examples of forward-looking statements include statements regarding the Enhanced Breast Cancer Detection (EBCD) program delivering equitable results for diverse populations across different racial, ethnic and breast density patient groups, and are qualified by the inherent risks and uncertainties surrounding future expectations generally, all of which are subject to change. Actual results could differ materially from those currently anticipated due to a number of risks and uncertainties, many of which are beyond RadNet’s control.

Forward-looking statements are neither historical facts nor assurances of future performance. Instead, they are based only on management’s current beliefs, expectations and assumptions regarding the future of RadNet’s business, future plans and strategies, projections, anticipated events and trends, the economy and other future conditions. Because forward-looking statements relate to the future, they are subject to inherent uncertainties, risks and changes in circumstances that are difficult to predict and many of which are outside of RadNet’s control. RadNet’s actual results and financial condition may differ materially from those indicated in the forward-looking statements as a result of various factors. Neither RadNet nor its directors, executive officers, or advisors, provide any representation, assurance or guarantee that the occurrence of the events expressed or implied in any forward-looking statements will actually occur, or if any of them do occur, what impact they will have on the business, results of operations or financial condition of RadNet. Should any risks and uncertainties develop into actual events, these developments could have a material adverse effect on RadNet’s business and the ability to realize the expected benefits of the acquisition. Risks and uncertainties that could cause results to differ from expectations include, but are not limited to: (1) the ability to recognize the anticipated benefits of the technology, and (2) the risk of legislative, regulatory, economic, competitive, and technological changes. The foregoing review of important factors should not be construed as exhaustive and should be read in conjunction with the other cautionary statements that are included elsewhere. Additional information concerning risks, uncertainties and assumptions can be found in RadNet’s filings with the Securities and Exchange Commission (the “SEC”), including the risk factors discussed in RadNet’s most recent Annual Report on Form 10-K, as updated by its Quarterly Reports on Form 10-Q and future filings with the SEC.

Forward-looking statements included herein are made only as of the date hereof and, except as required by applicable law, RadNet does not undertake any obligation to update any forward-looking statements, or any other information in this communication, as a result of new information, future developments or otherwise, or to correct any inaccuracies or omissions in them which become apparent. All forward-looking statements in this communication are qualified in their entirety by this cautionary statement.

Media Contact
Jane Mazur
SVP, Corporate Communications
RadNet
+1 585-355-5978
jane.mazur@radnet.com

Mark Stolper
Executive Vice President and Chief Financial Officer
RadNet
+1 310-445-2800

References

1)   Based on a review of all PubMed results as of November 2025 from the search query: (mammography) AND (AI) AND (cancer detection rate OR real world OR United States).

2)   Seidenwurm D, Lee CS, Bhargavan-Chatfield M, et al. “Assessing the Recall Rate for Screening Mammography: Comparing the Medicare Hospital Compare Dataset With the National Mammography Database.” American Journal of Roentgenology. 2018.

3)   “Breast cancer death rates are highest for Black women—again.” American Cancer Society. October 3, 2022.

4)   “About Dense Breasts.” Centers for Disease Control and Prevention. September 11, 2024.

5)   Data on file. RadNet.

6)   Kim et al. “Impact of a Categorical AI System for Digital Breast Tomosynthesis on Breast Cancer Interpretation by Both General Radiologists and Breast Imaging Specialists.” Radiology: Artificial Intelligence. February 7, 2024.

FAQ

What did RadNet (RDNT) announce on Nov 17, 2025 about DeepHealth AI and breast cancer screening?

RadNet announced ASSURE study results showing DeepHealth’s AI workflow increased cancer detection by 21.6% vs 3D mammography in a >579,000-woman real-world analysis.

How did DeepHealth’s AI affect detection for women with dense breasts in the ASSURE study (RDNT)?

The AI-enabled workflow delivered a 22.7% higher cancer detection rate for women with dense breasts compared to 3D mammography.

Did the ASSURE study maintain recall rates within clinical guidelines for RadNet (RDNT)?

Yes; the study reported recall rates remained within American College of Radiology guidelines while improving detection and PPV.

How large and diverse was the patient population in RadNet’s ASSURE study (RDNT)?

The study used mammograms from >579,000 women across 109 community sites and included over 150,000 Black women.

What clinical benefit did DeepHealth’s Breast Suite provide in RadNet’s EBCD program?

DeepHealth’s Breast Suite powered an AI workflow that increased cancer detection, raised PPV by 15%, and enabled second-expert review for high-suspicion cases.