AI in Breast Cancer Screening: What the Research Actually Shows
Screening mammography AI, the oldest research base in the field and the one with the largest prospective trials.
Mammography AI has been studied longer than almost anything else here, and it carries a cautionary history worth knowing.
An earlier generation of computer-aided detection was rolled out widely in the United States, reimbursed, and then found in large reviews not to have improved outcomes the way everyone assumed. That is not a reason to dismiss the current generation, which works differently and performs better. It is a reason to insist on outcome data rather than accuracy data.
The recent European trials are the strongest evidence in the field, partly because those screening programs use double reading, which gives a genuinely hard comparison to beat.
What Kind of Evidence This Is
Not every study carries the same weight. This is the mix behind the list below.
What to Measure in Your Own Setting
A published result is somebody else's hospital. These are the numbers worth tracking in yours.
- Recall rate. Calling women back for extra imaging has a real cost in anxiety and in money, and it is the number that moved badly last time.
- Cancers detected per thousand screened, not sensitivity on a test set.
- Reader workload, if the plan is for software to replace one of two readers. That is where the savings are, and where the risk is.
- Interval cancers, the ones that appear between screens. They are the honest test of whether anything was actually missed.
Who Sells This, and What the FDA Has Cleared
The research above says whether the idea works. These are the two questions a buyer asks next.
Companies listed here doing this work
- Aidoc 30 FDA clearances on record Radiology triage AI that flags suspected acute findings such as intracranial hemorrhage and pulmonary embolism to move them up the read list. One of the larger FDA-cleared portfolios in imaging triage.
- Viz.ai 8 FDA clearances on record Care coordination built on imaging AI: detects suspected large vessel occlusion stroke and alerts the intervention team directly, compressing time-to-treatment rather than only flagging the scan.
- PathAI AI for digital pathology, applied to both diagnostic support and pharmaceutical research workflows where consistent slide interpretation is the bottleneck.
- Lunit 6 FDA clearances on record Imaging AI concentrated in mammography screening and chest radiography, with a significant deployment base in national screening programs outside the United States.
- Harrison.ai Radiology 8 FDA clearances on record Radiology AI formerly marketed as Annalise.ai, now operating under the Harrison.ai Radiology name across chest X-ray and CT decision support.
- RapidAI 14 FDA clearances on record Neurovascular imaging analysis for stroke and aneurysm workflows, widely embedded in comprehensive stroke center protocols.
- GE HealthCare 9 FDA clearances on record Public healthcare technology company spanning imaging, ultrasound, monitoring, diagnostics, and AI-enabled software used across hospital workflows.
- Siemens Healthineers 12 FDA clearances on record Public medical technology company with imaging, diagnostics, cancer-care, and digital businesses that use AI across equipment and clinical workflow.
- Philips Public health technology company using AI across diagnostic imaging, ultrasound, monitoring, informatics, and image-guided therapy.
- DeepHealth 10 FDA clearances on record RadNet subsidiary combining radiology informatics, workflow, and clinical AI across breast, chest, neuro, prostate, thyroid, and other imaging work.
- Avicenna.AI 7 FDA clearances on record CT software that flags urgent and incidental findings, sold as the CINA product family. It covers brain bleeds, large vessel occlusion, pulmonary embolism, aortic dissection and spine fractures, and is used by radiology departments and teleradiology providers.
- ScreenPoint Medical 7 FDA clearances on record Transpara reads 2D and 3D mammograms to help a radiologist find breast cancer, rate breast density, and compare a scan against a woman's earlier images. Screening programs are its main buyer.
- Brainomix 6 FDA clearances on record Brainomix 360 reads brain scans to support decisions in acute stroke care, and e-Lung measures features on lung CT scans. Hospitals use the stroke product in the treatment window, and drug companies use the lung product as a trial measurement.
- Qure.ai 4 FDA clearances on record Qure.ai reads chest X-rays, CT scans and other images to flag findings and move urgent cases up the list. The company says it holds 26 FDA-cleared indications across nine products, and it sells to hospitals, screening programs and public health projects in the United States, India and Africa.
- Therapixel 5 FDA clearances on record MammoScreen reads a mammogram and marks the places a radiologist should look again for breast cancer. Therapixel is a French software company, and breast screening programs and radiology groups are its buyers.
- Koios Medical 4 FDA clearances on record Koios DS reads ultrasound images of the breast and the thyroid and gives the radiologist a second opinion on whether a lump looks suspicious. The company says the software is FDA cleared, CE marked and a billable procedure.
- JLK 4 FDA clearances on record JLK sells MEDIHUB, a family of AI tools that read brain scans and other medical images, with stroke care as the main focus. The company is based in South Korea and its shares trade on the KOSDAQ market.
- AZmed 5 FDA clearances on record Rayvolve reads X-rays and marks fractures and chest findings for the radiologist. AZmed is based in Paris and sells to hospital radiology departments in Europe and the United States, and the company says Rayvolve now covers eight FDA-cleared findings.
- BrightHeart 3 FDA clearances on record The B-Right platform watches a prenatal ultrasound as it happens and helps the sonographer check that the baby's heart views are complete and read correctly. BrightHeart is based in Paris, and its two products, B-Right Views and B-Right Screen, are both FDA cleared and CE marked.
- Sonio 2 FDA clearances on record Sonio follows a fetal ultrasound as it happens, checks that the required views were actually captured, flags findings that look abnormal, and writes the report. The company says it has five FDA clearances in under three years, and Samsung now owns it.
- Odin Vision 3 FDA clearances on record CADDIE watches a colonoscopy as it happens and points out polyps the doctor should look at again, and CADU does the same job for a gastroscopy. Odin Vision is a London company, and Olympus bought it in 2023, so the software now sits beside Olympus endoscopy equipment.
- Hologic 6 FDA clearances on record Hologic makes breast imaging, surgical and laboratory products for women's health, and its Genius AI Detection software reads 3D mammograms and marks the spots a radiologist should check for cancer. Blackstone and TPG took the company private in April 2026, so its shares no longer trade.
- Digital Diagnostics 2 FDA clearances on record LumineticsCore looks at a photo of the back of the eye and decides on its own whether a person with diabetes has diabetic retinopathy, without an eye doctor reading the image. Primary care clinics are the main buyer, and the company says the result comes back in about 30 seconds.
- Nanox.AI 6 FDA clearances on record Reads routine CT scans a patient already had and flags signs of heart, liver, and bone conditions that were not the reason for the scan, so they can be caught earlier. Formerly Zebra Medical Vision.
Device types the FDA has cleared for this
- AI Cancer Detection Software for Radiology 40 cleared products Marks places on a scan that may be cancer, and in some cases gives a score for how suspicious they look.
- AI Diagnostic Software for Suspicious Lesions 23 cleared products Helps characterize a lesion that has already been found, rather than finding it in the first place.
The Studies
Every entry links to the source record. Summaries are the study's own abstract, shortened but not reworded.
-
International evaluation of an AI system for breast cancer screening.
Screening mammography aims to identify breast cancer at earlier stages of the disease, when treatment can be more successful1. Despite the existence of screening programmes worldwide, the interpretation of mammograms is affected by high rates of false positives and false negatives2. Here we present an artificial intelligence (AI) system that is capable of surpassing human experts in breast cancer prediction.
PMID 31894144 ... doi:10.1038/s41586-019-1799-6
-
A Deep Learning Mammography-based Model for Improved Breast Cancer Risk Prediction.
Background Mammographic density improves the accuracy of breast cancer risk models. However, the use of breast density is limited by subjective assessment, variation across radiologists, and restricted data. A mammography-based deep learning (DL) model may provide more accurate risk prediction.
PMID 31063083 ... doi:10.1148/radiol.2019182716
-
Screening mammography with computer-aided detection: prospective study of 12,860 patients in a community breast center.
PurposeTo prospectively assess the effect of computer-aided detection (CAD) on the interpretation of screening mammograms in a community breast center.Materials and methodsOver a 12-month period, 12,860 screening mammograms were interpreted with the assistance of a CAD system.
PMID 11526282 ... doi:10.1148/radiol.2203001282
-
Stand-Alone Artificial Intelligence for Breast Cancer Detection in Mammography: Comparison With 101 Radiologists.
BackgroundArtificial intelligence (AI) systems performing at radiologist-like levels in the evaluation of digital mammography (DM) would improve breast cancer screening accuracy and efficiency. We aimed to compare the stand-alone performance of an AI system to that of radiologists in detecting breast cancer in DM.MethodsNine multi-reader, multi-case study datasets previously used for different research purposes in se...
PMID 30834436 ... doi:10.1093/jnci/djy222
-
Diagnostic Accuracy of Digital Screening Mammography With and Without Computer-Aided Detection.
ImportanceAfter the US Food and Drug Administration (FDA) approved computer-aided detection (CAD) for mammography in 1998, and the Centers for Medicare and Medicaid Services (CMS) provided increased payment in 2002, CAD technology disseminated rapidly.
PMID 26414882 ... doi:10.1001/jamainternmed.2015.5231
-
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.
BackgroundRetrospective studies have shown promising results using artificial intelligence (AI) to improve mammography screening accuracy and reduce screen-reading workload; however, to our knowledge, a randomised trial has not yet been conducted.
PMID 37541274 ... doi:10.1016/s1470-2045(23)00298-x
-
Influence of computer-aided detection on performance of screening mammography.
BackgroundComputer-aided detection identifies suspicious findings on mammograms to assist radiologists. Since the Food and Drug Administration approved the technology in 1998, it has been disseminated into practice, but its effect on the accuracy of interpretation is unclear.MethodsWe determined the association between the use of computer-aided detection at mammography facilities and the performance of screening mamm...
PMID 17409321 ... doi:10.1056/nejmoa066099
-
Detection of Breast Cancer with Mammography: Effect of an Artificial Intelligence Support System.
Purpose To compare breast cancer detection performance of radiologists reading mammographic examinations unaided versus supported by an artificial intelligence (AI) system. Materials and Methods An enriched retrospective, fully crossed, multireader, multicase, HIPAA-compliant study was performed.
PMID 30457482 ... doi:10.1148/radiol.2018181371
-
Potential contribution of computer-aided detection to the sensitivity of screening mammography.
PurposeTo determine the false-negative rate in screening mammography, the capability of computer-aided detection (CAD) to identify these missed lesions, and whether or not CAD increases the radiologists' recall rate.Materials and methodsAll available screening mammograms that led to the detection of biopsy-proved cancer (n = 1,083) and the most recent corresponding prior mammograms (n = 427) were collected from 13 fa...
PMID 10796939 ... doi:10.1148/radiology.215.2.r00ma15554
-
Changes in cancer detection and false-positive recall in mammography using artificial intelligence: a retrospective, multireader study.
BackgroundMammography is the current standard for breast cancer screening. This study aimed to develop an artificial intelligence (AI) algorithm for diagnosis of breast cancer in mammography, and explore whether it could benefit radiologists by improving accuracy of diagnosis.MethodsIn this retrospective study, an AI algorithm was developed and validated with 170 230 mammography examinations collected from five insti...
PMID 33334578 ... doi:10.1016/s2589-7500(20)30003-0
-
Deep Learning to Improve Breast Cancer Detection on Screening Mammography.
The rapid development of deep learning, a family of machine learning techniques, has spurred much interest in its application to medical imaging problems. Here, we develop a deep learning algorithm that can accurately detect breast cancer on screening mammograms using an "end-to-end" training approach that efficiently leverages training datasets with either complete clinical annotation or only the cancer status (labe...
PMID 31467326 ... doi:10.1038/s41598-019-48995-4
-
Detecting and classifying lesions in mammograms with Deep Learning.
In the last two decades, Computer Aided Detection (CAD) systems were developed to help radiologists analyse screening mammograms, however benefits of current CAD technologies appear to be contradictory, therefore they should be improved to be ultimately considered useful. Since 2012, deep convolutional neural networks (CNN) have been a tremendous success in image recognition, reaching human performance.
PMID 29545529 ... doi:10.1038/s41598-018-22437-z
-
Evaluation of Combined Artificial Intelligence and Radiologist Assessment to Interpret Screening Mammograms.
ImportanceMammography screening currently relies on subjective human interpretation. Artificial intelligence (AI) advances could be used to increase mammography screening accuracy by reducing missed cancers and false positives.ObjectiveTo evaluate whether AI can overcome human mammography interpretation limitations with a rigorous, unbiased evaluation of machine learning algorithms.Design, setting, and participantsIn...
PMID 32119094 ... doi:10.1001/jamanetworkopen.2020.0265
-
A curated mammography data set for use in computer-aided detection and diagnosis research.
Published research results are difficult to replicate due to the lack of a standard evaluation data set in the area of decision support systems in mammography; most computer-aided diagnosis (CADx) and detection (CADe) algorithms for breast cancer in mammography are evaluated on private data sets or on unspecified subsets of public databases.
PMID 29257132 ... doi:10.1038/sdata.2017.177
-
Mammographic characteristics of 115 missed cancers later detected with screening mammography and the potential utility of computer-aided detection.
PurposeTo retrospectively determine the mammographic characteristics of cancers missed at screening mammography and assess the ability of computer-aided detection (CAD) to mark the missed cancers.Materials and methodsA multicenter retrospective study accrued 1,083 consecutive cases of breast cancer detected at screening mammography. Prior mammograms were available in 427 cases.
PMID 11274556 ... doi:10.1148/radiology.219.1.r01ap16192
-
Contrast-enhanced Mammography: State of the Art.
Contrast-enhanced mammography (CEM) has emerged as a viable alternative to contrast-enhanced breast MRI, and it may increase access to vascular imaging while reducing examination cost. Intravenous iodinated contrast materials are used in CEM to enhance the visualization of tumor neovascularity.
PMID 33650905 ... doi:10.1148/radiol.2021201948
-
Robust breast cancer detection in mammography and digital breast tomosynthesis using an annotation-efficient deep learning approach.
Breast cancer remains a global challenge, causing over 600,000 deaths in 2018 (ref. 1). To achieve earlier cancer detection, health organizations worldwide recommend screening mammography, which is estimated to decrease breast cancer mortality by 20-40% (refs. 2,3).
PMID 33432172 ... doi:10.1038/s41591-020-01174-9
-
Automation Bias in Mammography: The Impact of Artificial Intelligence BI-RADS Suggestions on Reader Performance.
Background Automation bias (the propensity for humans to favor suggestions from automated decision-making systems) is a known source of error in human-machine interactions, but its implications regarding artificial intelligence (AI)-aided mammography reading are unknown.
PMID 37129490 ... doi:10.1148/radiol.222176
-
Deep Learning in Mammography: Diagnostic Accuracy of a Multipurpose Image Analysis Software in the Detection of Breast Cancer.
ObjectivesThe aim of this study was to evaluate the diagnostic accuracy of a multipurpose image analysis software based on deep learning with artificial neural networks for the detection of breast cancer in an independent, dual-center mammography data set.Materials and methodsIn this retrospective, Health Insurance Portability and Accountability Act-compliant study, all patients undergoing mammography in 2012 at our ...
PMID 28212138 ... doi:10.1097/rli.0000000000000358
-
Artificial intelligence for breast cancer detection in screening mammography in Sweden: a prospective, population-based, paired-reader, non-inferiority study.
BackgroundArtificial intelligence (AI) as an independent reader of screening mammograms has shown promise, but there are few prospective studies. Our aim was to conduct a prospective clinical trial to examine how AI affects cancer detection and false positive findings in a real-world setting.MethodsScreenTrustCAD was a prospective, population-based, paired-reader, non-inferiority study done at the Capio Sankt Göran H...
PMID 37690911 ... doi:10.1016/s2589-7500(23)00153-x
-
Effect of artificial intelligence-based triaging of breast cancer screening mammograms on cancer detection and radiologist workload: a retrospective simulation study.
BackgroundWe examined the potential change in cancer detection when using an artificial intelligence (AI) cancer-detection software to triage certain screening examinations into a no radiologist work stream, and then after regular radiologist assessment of the remainder, triage certain screening examinations into an enhanced assessment work stream.
PMID 33328114 ... doi:10.1016/s2589-7500(20)30185-0
-
External Evaluation of 3 Commercial Artificial Intelligence Algorithms for Independent Assessment of Screening Mammograms.
ImportanceA computer algorithm that performs at or above the level of radiologists in mammography screening assessment could improve the effectiveness of breast cancer screening.ObjectiveTo perform an external evaluation of 3 commercially available artificial intelligence (AI) computer-aided detection algorithms as independent mammography readers and to assess the screening performance when combined with radiologists...
PMID 32852536 ... doi:10.1001/jamaoncol.2020.3321
-
Artificial Intelligence for Mammography and Digital Breast Tomosynthesis: Current Concepts and Future Perspectives.
Although computer-aided diagnosis (CAD) is widely used in mammography, conventional CAD programs that use prompts to indicate potential cancers on the mammograms have not led to an improvement in diagnostic accuracy.
PMID 31549948 ... doi:10.1148/radiol.2019182627
-
Toward robust mammography-based models for breast cancer risk.
Improved breast cancer risk models enable targeted screening strategies that achieve earlier detection and less screening harm than existing guidelines. To bring deep learning risk models to clinical practice, we need to further refine their accuracy, validate them across diverse populations, and demonstrate their potential to improve clinical workflows.
PMID 33504648 ... doi:10.1126/scitranslmed.aba4373
Study records come from Europe PMC, which indexes PubMed, MEDLINE, PMC and preprint servers. Titles, journals, years, identifiers, citation counts and abstracts are reproduced from the source record and are not rewritten here. Listing a study is not an endorsement of its conclusion, and citation count measures attention rather than quality. This page is a starting point for your own reading, not clinical guidance.