1. Introduction

Breast cancer (BC) remains the most frequently diagnosed malignancy in women, being ranked first in terms of cancer-related fatality for women all around the world (1). The current standards in breast cancer care underscore the importance of a comprehensive, multidisciplinary approach that implies multiple phases: prevention, early detection, diagnosis, treatment, and prognosis. The American College of Surgeons’ National Accreditation Program for Breast Centers has recently outlined the “Optimal Resources for Breast Care (2024 Standards)”, concentrating on the patient care journey starting with screening and prevention to diagnosis, treatment, and prognosis, with an emphasis on the importance of providing value-based care with multidisciplinary collaboration (2).

Breast cancer care has changed significantly, due to technological advances and continuous research. However, several challenges persist and continue to bring several dysfunctionalities in this field, affecting patient outcomes. A notable issue that causes this is the disparity in access to healthcare services, which is responsible for late-stage diagnoses and even delays in receiving the appropriate treatment. These challenges are mainly accentuated in low- and middle-income countries, where healthcare infrastructure might be inadequate (3,4).

Medical imaging is of the utmost importance in the assessment of BC nowadays, mammography and ultrasonography representing the primary two methods for screening, therapy guidance (surgically or percutaneously, in the case of vacuum excisions, cryoablation, or radiofrequency ablation), and assisting tissue sampling in the case of imaging-guided biopsies. Magnetic Resonance Imaging (MRI) shows the highest sensitivity in detecting BC, but it has a high percentage of false positive results. Unless lesions indicated via MRI are also seen using ultrasonography or mammography, a biopsy is required. It is realized by MRI guidance, since it is the only suitable method in that case (5).

Despite the significant advancements in breast cancer care over the last decades, several challenges remain unresolved. Human subjectivity in interpreting diagnostic results is one of the most challenging aspects of it, because it can adversely impair the accuracy of diagnosis. The persistence of false-negative results in imaging examinations continues to influence the whole management of the case and affect patient outcomes (6).

Artificial intelligence (AI) is defined Cambridge Dictionary as “the use or study of computer systems or machines that have some of the qualities that the human brain has, such as the ability to interpret and produce language in a way that seems human, recognize or create images, solve problems, and learn from data supplied to them”.

AI can be viewed as a transformative force in breast cancer care, due to its abilities to analyze complex datasets, recognize patterns, and make predictions that are capable of surpassing human capabilities in speed and precision. One of its most significant contributions is seen in diagnostic imaging, where AI-powered processes (such as deep learning algorithms, which are specific to AI) excel at interpreting variable imaging examinations, including mammograms, ultrasounds, and MRIs (7).

The objective of this review is to present the potential of AI to transform breast imaging in terms of improving diagnostic accuracy. We are also debating the need for AI-assisted decisions, focusing on how these technologies supplement the decision-making process, particularly in complex cases where human subjectivity may influence outcomes.

2. Materials and Methods

We used the keywords “Artificial Intelligence” AND “Breast Cancer”, filtering for results published from 2020 to 2025. To be included, articles had to be available in English and focus on the applications of AI programs in the field of breast cancer imaging, including screening and diagnosis.

We excluded studies on animals, abstracts, and articles lacking sufficient details on methods and results. Two authors screened the titles and abstracts and further analyzed the full text of the selected articles. We identified the recurring themes, types of AI algorithms used, areas of application, and reported benefits or limitations. The synthesis was performed using a narrative approach.

3. Results

We selected 13 comprehensive reviews and one systematic review for our analysis, along with three original studies (Tables 1 and 2).

Table 1. Reviews included in our analysis (all are narrative reviews except for one, marked with * which is a systematic review)

AI – artificial intelligence, DL – deep learning, ML – machine learning, CNNs – convolutional neural networks, DCE-MRI – dynamic contrast-enhanced magnetic resonance imaging

Table 2. Original research studies included in the review

DBT – digital breast tomosynthesis, DL – deep learning, ML – machine learning, LLM – large language model, PPV – positive predictive value, NPV – negative predictive value, R0 complete excision, R1 – incomplete excision.

Deep learning models have been shown to possess a remarkable ability that allows the identification of patterns in imaging data that a human examiner could miss. Some studies have demonstrated that AI can help in the detection of early-stage tumors in mammograms, ultrasounds, and MRIs, reducing the rate of false negative results and enabling earlier intervention (8,9).

Furthermore, many clinical decisions can be assisted by AI (clinical decision support systems) and applied in patients with breast pathology, from diagnosis to treatment (9,10).

3.1. AI application for diagnosis

AI-assisted programs have autonomy and are in a constant self-development process that leads to increasing the accuracy and speed of detection (11,12). Occult lesions were shown to be detected more commonly using such AI programs, findings that could lead to treatments that are initiated earlier, not only improving the survival rates, but also reducing the need for more aggressive therapies later (5).

3.1.1. Mammography

AI algorithmic methods integrated into mammography enable the extraction of texture and intensity-based features such as entropy, skewness, and variance, facilitating accurate classification using machine learning (ML) models.

Even in dense breasts, where traditional detection using mammography is challenging, AI improves diagnostic performance. Advanced models achieved detection accuracy up to 98.96%, sensitivity of 100%, and a specificity of 94%. Segmentation methods using dense U-Nets with attention gates reached an accuracy of 78.38% and Mask R-CNN with G-CNN achieved 99.01% accuracy and 99.24% sensitivity (7).

The AI performance for breast cancer detection increases when clinical data is integrated, according to Trang et al. Combining mammography images and clinical records on 731 images from 357 women and using deep CNN models and ML classifiers achieved 84.5% accuracy, 89.7% sensitivity, 78.1% specificity, and an AUC of 0.88. The combined model outperformed models using mammograms alone, which led to an accuracy of 72.5% (8).

A systematic review by Rentiya et al. analyzed 55 studies showing that AI improves breast cancer care. AI models achieved high diagnostic accuracy in mammography (AUC up to 0.92) (10).

Similar values were reported in the review of Gao et al., which reported values ranging from 68% to 99%, with some approaches achieving AUC values exceeding 0.98 (12).

AI is already integrated into commercially available medical devices. A study by Carriero et al. analyzed multiple AI detection platforms for mammography, showing high accuracy and efficiency while reducing false positives and negatives (13).

When included in digital mammography and digital breast tomosynthesis analysis, AI had to analyze a higher number of images than in 2D mammography. However, it still showed a performance comparable to or even better than that of radiologists, with increased sensitivity, specificity, and AUC (14).

Letter et al. (2024) conducted a multicenter retrospective study that included 13,885 mammography screening examinations using digital tomosynthesis (DBT), of which 5,883 were analyzed using the iCAD ProFound AI v2.0 system and 7,002 without the use of AI. The aim was to evaluate the real-world impact of this FDA-approved tool on key screening indicators (cancer detection rate (CDR) and abnormal interpretation rate (AIR)) in current clinical practice. Even though no statistically significant differences were identified, trends toward increased early cancer detection and biopsy indication accuracy were observed when AI was used (15).

3.1.2. Ultrasound

AI was also tested on ultrasound-generated images. In their review, Li et al. included nine original research studies focused on the AI implementation for ultrasound, of which four were focused on breast tumor identification or classification. The AI achieved a higher AUC than the average of ten breast radiologists (0.962 versus 0.924) in one of the studies. An accuracy of 95.3% compared with the accuracy of 94.1% on human readers was reached for differentiating BI-RADS 2–3 versus BI-RADS 4–5 by a CNN model. (16).

3.2. AI impact on radiology workflow

AI enables improvements in breast imaging by simplifying the workflow, automating monotonous tasks, and reducing reading times (5, 9, 11). This leads to better time management for medical specialists, allowing them to focus on complex or suspicious lesions (9). With the accelerated advancement of computer science, AI is thought to have significant advantages in processing image data (17).

Radiologist workload decreased using AI, due to its performance in excluding normal images, as concluded by the authors of the above-mentioned systematic review (10). Similarly, Guo et al. acknowledge that ML techniques in breast cancer imaging can improve diagnostic accuracy and efficiency, leading to better outcomes and lower costs (18). Time reduction by up to 50% and decreasing workload through triaging normal cases was mentioned also by Ahn JS et al, in their comprehensive review (14).

Another field in which AI can bring a contribution is determining personalized screening intervals using breast density as a biomarker. This aspect is important, taking into account that MRI costs ten times more than mammography but provides a higher negative predictive value, reducing unnecessary biopsies (5).

Although combined use with radiologists improves detection rates, AI systems require careful evaluation to ensure consistent and effective integration into clinical practice (11). ML depends on extensive, high-quality data, cannot diagnose untrained diseases, and faces challenges with data privacy and generalizability (18).

3.3. Other applications

Automatic image analysis can help intraoperatory decisions by identifying close or positive margins on surgical specimens. Andras et al. conducted a retrospective study on 100 patients, in which ChatGPT-4 achieved 84% accuracy in predicting surgical margin status on intraoperative mammograms. The non-specialized AI showed 60% sensitivity for R1 (incomplete excision), 86.7% specificity for R0 (complete excision), a positive predictive value (PPV) of 33.3%, NPV 95.1%, F1 score 0.43, and Cohen’s kappa 0.34, indicating moderate agreement with histopathology (19). Much better performance would probably be achieved by specialized models.

Moving beyond tumor detection, AI can identify molecular subtypes using macro images such as MRI, addressing biopsy limitations in capturing tumor heterogeneity. A review by Ameen et al. discussed how DCE-MRI enables accurate, non-invasive prediction of breast cancer molecular subtypes. Studies using CNN and ML models on post-contrast T1, ADC, and T2 sequences achieved high AUCs (up to 0.920), with best results in identifying HER2-enriched and triple-negative cancers (20).

AI also effectively predicts axillary lymph node metastasis, therapy response, and prognosis, outperforming traditional methods (16). AI-enhanced breast cancer risk models integrate imaging, clinical, genetic, and pathology data, enabling personalized screening and prevention. CNNs improved pCR prediction in HER2+ cases and the results were statistically superior to traditional prediction methods, according to a study included in the review by You C. et al (17). New imaging methods and genetic testing improve model accuracy and emphasize that a multidisciplinary approach is key for earlier detection and improved outcomes (21).

 3.4. Radiomics

Continuously gaining popularity, radiomics is an innovative domain that refers to the characteristics of the tumor that cannot be detected by the human eye, but can show microstructural patterns that can be viewed with the use of imaging techniques., By its capacity to identify minimal lesions, ML can contribute to the development of radiomics, decreasing the rate of false-negative results and improving the management of the case (20,22).

According to Tagliafico et al., ML enables automated patient stratification and outcome prediction using imaging data. Unsupervised ML clusters patient data, while supervised ML trains on labeled datasets to predict disease presence or progression. Advanced ML methods like Random Forests and regularization networks rank features for predictive power. Deep learning (AI) with CNNs automates feature extraction directly from images, enhancing multi-modal analysis, and positions AI as a leading tool for next-generation radiomics in breast cancer (22).

4. Discussion

In this review, we briefly discussed recent data regarding applications and performance of AI-based solutions for breast cancer imaging. The AI models have show high accuracy in image interpretation, which can be used as first reader and speed up the process for screening and diagnosis. Implementing such solutions positively impacts workflow by reducing the workload. Complex image analysis, including radiomics, is currently under development for molecular types identification and predicting response to systemic therapies.

Integrating clinical data into AI models increased the accuracy of imaging diagnosis. Taking into account associated pathologies and personal preference of the patient allows a better integration of AI in personalized patient care (18,22,23). At least a basic understanding of the functionality, advantages, and drawbacks of these algorithms is essential to integrate AI models into clinical practice responsibly (19,25,26).

The use of extensive datasets by AI algorithms raises significant concerns about patient data safety and confidentiality. Increasing the quality of AI systems leads to better performance in differentiating valid information from inadequate or erroneous data (19,27,28). However, the ethical development and evaluation of AI algorithms in medicine need to prioritize the fundamental principles of nonmaleficence and beneficence, especially when considering patient safety.

Despite the current encouraging results, further clinical validation through large-scale multi-center validation studies, including various patient populations, is required to ensure consistent performance across different clinical settings. These advancements may support the transition of AI from pilot applications to standardized clinical tools, contributing to earlier diagnosis, personalized treatment planning, and improved outcomes in breast cancer care.

Our review has some limitations. The main one is the risk of bias that could have occurred in the article selection process. Another limitation lies in the semi-structured approach and the level of scientific evidence provided by the narrative reviews, which were the primary material for the analysis.

5. Conclusion

All articles included in this review consistently highlighted the potential utility of AI for breast cancer imaging. AI tools were proven beneficial for improving diagnostic accuracy and enhancing workflow efficiency. Consistent results in various clinical settings and population need to be supported by further research.

ABBREVIATIONS

AI – Artificial intelligence

AIR – Abnormal Interpretation Rate

AUC – Area Under the Curve

BC – Breast cancer

BI-RADS – Breast Imaging-Reporting and Data System

CDR – Cancer Detection Rate

CNN – Convolutional Neural Network

DBT – Digital Breast Tomosynthesis

DCE-MRI – Dynamic contrast-enhanced Magnetic Resonance Imaging

HER2 – Human Epidermal Growth Factor Receptor 2

ML – Machine Learning

MRI – Magnetic Resonance Imaging

NPV – Negative Predictive Value

pCR – Pathologic complete response

PPV – Positive Predictive Value

STATEMENTS

Author Contributions: Conceptualization, RAI, DA, VE, and GD; methodology, DA, RAI, VE, and AC; formal analysis, RAI, VE, AI, and MC; investigation, RAI, DA, AI, and MC; data curation, RAI, VE and MC; writing—original draft preparation, RAI and MC; writing—review and editing, DA, VE, AI, and AC; supervision GD and AC. All authors have read and agreed to the published version of the manuscript.

Funding: The current work was conducted as part of the clinical study entitled “Improving Postoperative Outcomes in Breast Surgery through the Use of Artificial Intelligence and Augmented Reality Programs”, carried out at the Cluj-Napoca County Emergency Clinical Hospital, 1st Surgery Department, with the support of the Iuliu Hațieganu University of Medicine and Pharmacy, Cluj-Napoca. This research received no external funding.

Conflicts of Interest: The authors declare no conflicts of interest.

REFERENCES

1. Breast Cancer (Internet). World Health Organization. (cited 10 August 2025). Available online: https://www.who.int/news-room/fact-sheets/detail/breast-cancer.
2. American College of Surgeons. NAPBC standards and resources (Optimal Resources for Breast Care, 2024 Standards) (Internet). American College of Surgeons; 2024 (cited 2025 Aug 11). Available from: https://www.facs.org/quality-programs/cancer-programs/national-accreditation-program-for-breast-centers/standards-and-resources/ ACS
3. Barrios CH. Global challenges in breast cancer detection and treatment. Breast. 2022 Mar;62 Suppl 1(Suppl 1):S3-S6. doi: 10.1016/j.breast.2022.02.003. Epub 2022 Feb 8. PMID: 35219542; PMCID: PMC9097801.
4. McAlearney AS, Murray K, Sieck C, Lin JJ, Bellacera B, Bickell NA. The Challenge of Improving Breast Cancer Care Coordination in Safety-net Hospitals: Barriers, Facilitators, and Opportunities. Med Care. 2016 Feb;54(2):147-54. doi: 10.1097/MLR.0000000000000458. PMID: 26565530; PMCID: PMC4713342.
5. Baltzer PAT. Künstliche Intelligenz in der Mammadiagnostik : Anwendungsgebiete aus klinischer Perspektive (Artificial intelligence in breast imaging : Areas of application from a clinical perspective). Radiologe. 2021 Feb;61(2):192-198. German. doi: 10.1007/s00117-020-00802-2. Epub 2021 Jan 28. PMID: 33507318; PMCID: PMC7851036.
6. Thekkinkattil D, Vidya R, Kwong A, Alajmi AA, Mutebi M, Gulluoglu B, Deo S, Fukuma E, Elder E, Gonzalez E, Warnberg F, Buccimazza I, Ung O, Walker M, Vernet-Tomas M, Peeters MTV, Johnson N, Paulinelli RR, Kuehn T, Veronesi P, Sarkar D, Dietz J. Assessing variability in breast cancer management across the world: results of a questionnaire survey amongst global international experts in breast cancer management. Ecancermedicalscience. 2022 Sep 2;16:1443. doi: 10.3332/ecancer.2022.1443. PMID: 36405941; PMCID: PMC9666278.
7. Al-Karawi D, Al-Zaidi S, Helael KA, Obeidat N, Mouhsen AM, Ajam T, Alshalabi BA, Salman M, Ahmed MH. A Review of Artificial Intelligence in Breast Imaging. Tomography. 2024 May 9;10(5):705-726. doi: 10.3390/tomography10050055. PMID: 38787015; PMCID: PMC11125819.
8. Trang NTH, Long KQ, An PL, Dang TN. Development of an Artificial Intelligence-Based Breast Cancer Detection Model by Combining Mammograms and Medical Health Records. Diagnostics (Basel). 2023 Jan 17;13(3):346. doi: 10.3390/diagnostics13030346. PMID: 36766450; PMCID: PMC9913958.
9. Díaz O, Rodríguez-Ruíz A, Sechopoulos I. Artificial Intelligence for breast cancer detection: Technology, challenges, and prospects. Eur J Radiol. 2024 Jun;175:111457. doi: 10.1016/j.ejrad.2024.111457. Epub 2024 Apr 16. PMID: 38640824.
10. Rentiya ZS, Mandal S, Inban P, Vempalli H, Dabbara R, Ali S, Kaur K, Adegbite A, Intsiful TA, Jayan M, Odoma VA, Khan A. Revolutionizing Breast Cancer Detection With Artificial Intelligence (AI) in Radiology and Radiation Oncology: A Systematic Review. Cureus. 2024 Apr 4;16(4):e57619. doi: 10.7759/cureus.57619. PMID: 38711711; PMCID: PMC11073588.
11. Dileep G, Gianchandani Gyani SG. Artificial Intelligence in Breast Cancer Screening and Diagnosis. Cureus. 2022 Oct 15;14(10):e30318. doi: 10.7759/cureus.30318. PMID: 36381716; PMCID: PMC9650950.
12. Gao Y, Lin J, Zhou Y, Lin R. The application of traditional machine learning and deep learning techniques in mammography: a review. Front Oncol. 2023 Aug 11;13:1213045. doi: 10.3389/fonc.2023.1213045. PMID: 37637035; PMCID: PMC10453798.
13. Carriero A, Groenhoff L, Vologina E, Basile P, Albera M. Deep Learning in Breast Cancer Imaging: State of the Art and Recent Advancements in Early 2024. Diagnostics (Basel). 2024 Apr 19;14(8):848. doi: 10.3390/diagnostics14080848. PMID: 38667493; PMCID: PMC11048882.
14. Ahn JS, Shin S, Yang SA, Park EK, Kim KH, Cho SI, Ock CY, Kim S. Artificial Intelligence in Breast Cancer Diagnosis and Personalized Medicine. J Breast Cancer. 2023 Oct;26(5):405-435. doi: 10.4048/jbc.2023.26.e45. PMID: 37926067; PMCID: PMC10625863.
15. Letter H, Peratikos M, Toledano A, Hoffmeister J, Nishikawa R, Conant E, Shisler J, Maimone S, Diaz de Villegas H. Use of Artificial Intelligence for Digital Breast Tomosynthesis Screening: A Preliminary Real-world Experience. J Breast Imaging. 2023 May 22;5(3):258-266. doi: 10.1093/jbi/wbad015. PMID: 38416890.
16. Li JW, Sheng DL, Chen JG, You C, Liu S, Xu HX, Chang C. Artificial intelligence in breast imaging: potentials and challenges. Phys Med Biol. 2023 Nov 28;68(23). doi: 10.1088/1361-6560/acfade. PMID: 37722385.
17. You C, Shen Y, Sun S, Zhou J, Li J, Su G, Michalopoulou E, Peng W, Gu Y, Guo W, Cao H. Artificial intelligence in breast imaging: Current situation and clinical challenges. Exploration (Beijing). 2023 Jul 20;3(5):20230007. doi: 10.1002/EXP.20230007. PMID: 37933287; PMCID: PMC10582610.
18. Guo Y, Zhang H, Yuan L, Chen W, Zhao H, Yu QQ, Shi W. Machine learning and new insights for breast cancer diagnosis. J Int Med Res. 2024 Apr;52(4):3000605241237867. doi: 10.1177/03000605241237867. PMID: 38663911; PMCID: PMC11047257.
19. Andras D, Ilies RA, Esanu V, Agoston S, Marginean Jumate TF, Dindelegan GC. Artificial Intelligence as a Potential Tool for Predicting Surgical Margin Status in Early Breast Cancer Using Mammographic Specimen Images. Diagnostics (Basel). 2025 May 17;15(10):1276. doi: 10.3390/diagnostics15101276. PMID: 40428269; PMCID: PMC12109882.
20. Ameen A, Shaikh K, Khan A, Vohra LM. The discerning influence of dynamic contrast-enhanced MRI in anticipating molecular subtypes of breast cancer through the artistry of artificial intelligence – a narrative review. J Pak Med Assoc. 2024 Apr;74(4 (Supple-4)):S72-S78. doi: 10.47391/JPMA.AKU-9S-11. PMID: 38712412.
21. Pesapane F, Battaglia O, Pellegrino G, Mangione E, Petitto S, Fiol Manna ED, Cazzaniga L, Nicosia L, Lazzeroni M, Corso G, Fusco N, Cassano E. Advances in breast cancer risk modeling: integrating clinics, imaging, pathology and artificial intelligence for personalized risk assessment. Future Oncol. 2023 Dec;19(38):2547-2564. doi: 10.2217/fon-2023-0365. Epub 2023 Dec 12. PMID: 38084492.
22. Tagliafico AS, Piana M, Schenone D, Lai R, Massone AM, Houssami N. Overview of radiomics in breast cancer diagnosis and prognostication. Breast. 2020 Feb;49:74-80. doi: 10.1016/j.breast.2019.10.018. Epub 2019 Nov 6. PMID: 31739125; PMCID: PMC7375670.
23. Farhud DD, Zokaei S. Ethical Issues of Artificial Intelligence in Medicine and Healthcare. Iran J Public Health. 2021 Nov;50(11):i-v. doi: 10.18502/ijph.v50i11.7600. PMID: 35223619; PMCID: PMC8826344.
24. MacIntyre MR, Cockerill RG, Mirza OF, Appel JM. Ethical considerations for the use of artificial intelligence in medical decision-making capacity assessments. Psychiatry Res. 2023 Oct;328:115466. doi: 10.1016/j.psychres.2023.115466. Epub 2023 Sep 7. PMID: 37717548.
25. Inglada Galiana L, Corral Gudino L, Miramontes González P. Ethics and artificial intelligence. Rev Clin Esp (Barc). 2024 Mar;224(3):178-186. doi: 10.1016/j.rceng.2024.02.003. Epub 2024 Feb 12. PMID: 38355097.
26. Sung J. Artificial intelligence in medicine: Ethical, social and legal perspectives. Ann Acad Med Singap. 2023 Dec 28;52(12):695-699. doi: 10.47102/annals-acadmedsg.2023272. PMID: 38920162.
27. Astărăstoae V, Rogozea LM, Leaşu F, Ioan BG. Ethical Dilemmas of Using Artificial Intelligence in Medicine. Am J Ther. 2024 Jul-Aug 01;31(4):e388-e397. doi: 10.1097/MJT.0000000000001693. Epub 2024 Apr 25. PMID: 38662923.
28. Tang L, Li J, Fantus S. Medical artificial intelligence ethics: A systematic review of empirical studies. Digit Health. 2023 Jul 6;9:20552076231186064. doi: 10.1177/20552076231186064. PMID: 37434728; PMCID: PMC10331228.