Breast cancer is a global health problem. In 2022, 2.3 million women were diagnosed, and 670,000 died from the disease. In Australia, it’s the most common cancer in women, 28% of all new cancer cases. The lifetime risk for Australian women is 1 in 7, with 21,194 new cases expected in 2024. Despite improvements in screening and treatment, the increasing incidence of breast cancer is putting huge pressure on the healthcare system. Pathologists, especially those in breast cancer diagnosis, are working longer hours, leading to delays in diagnosis and treatment. This is where AI can help in breast cancer detection, improve diagnosis and ease the burden on healthcare professionals.
Mammography and Pathology: A Team Effort in Breast Cancer Diagnosis
Mammography is the first step in breast cancer detection, providing radiologists with X-ray images of the breast tissue to identify potential abnormalities. While mammography is good at highlighting areas of concern, lumps or calcifications, it can’t confirm whether these are cancerous. This is where pathology comes in.
When a suspicious area is identified through mammography, a biopsy is often performed to get a tissue sample, which is then sent to a pathologist to examine. Pathologists examine the tissue under the microscope to confirm if the cells are cancerous or benign. The collaboration between mammography and pathology is key to an accurate diagnosis. Mammography can suggest potential issues, and pathology provides the definitive answer and detailed analysis for treatment decisions.
Cancer detection AI can help both processes by identifying suspicious areas in mammograms, aiming to increase detection rates without increasing false positives. If a biopsy is required, AI can also help pathologists by providing more information on the tissue, reducing workload and enabling more accurate diagnoses. Mammography, pathology and AI work together to improve early breast cancer detection and outcomes.
Mammography with AI
Mammography is the standard screening method for breast cancer. However, it has limitations, including breast tissue density and variability in the interpretation of images by different radiologists. AI algorithms have been developed to enhance mammographic image analysis, to identify subtle patterns that could indicate cancer. These AI systems trained on a large dataset act as a second reader to help pathologists detect areas of concern that might be missed. The AI model’s performance, as shown in various clinical studies can improve sensitivity and reduce false negatives.
In a German study, AI integrated into mammography screenings increased cancer detection by 17.6% without increasing false positive rates. AI can pinpoint specific areas of concern with confidence so pathologists are alerted to findings that need closer examination.
Global AI Implementation and Breast Cancer Trials
Several countries are running trials to evaluate AI in breast cancer screening. The UK’s National Health Service (NHS) is running a large trial with 700,000 mammograms, preliminary results show AI can reduce radiologist workload without compromising diagnostic performance. These trials demonstrate the potential of AI to improve the efficiency and accuracy of breast cancer diagnostics in pathology, making it an essential tool in the future of cancer detection.
In Sweden, AI has been integrated into the national screening program and has shown positive results in both diagnostic workflows and patient management. These early implementations show AI can support pathologists rather than replace them by acting as a reliable second reader in breast cancer diagnosis.
AI for Breast Cancer Risk Prediction and Personalised Screening
AI’s application in breast cancer detection goes beyond image analysis. It’s also being used to predict individual breast cancer risk by combining imaging data with patient history. This allows AI to stratify patients based on their risk, so that more targeted and efficient screening can be performed. By using AI to identify high-risk patients, healthcare providers can ensure those who need closer monitoring receive the care they need and optimise resource use.
There are also developments suggesting AI may be able to not only identify cancer that is present in a sample today, but also predict cancer in the future. MIT researchers have developed an AI model that can predict breast cancer within 5 years from mammographic data. This kind of predictive capability could be game-changing in delivering personalised screening and early intervention.
Challenges and Considerations in Clinical Adoption
While AI has great potential in breast cancer detection and pathology, there are challenges in its clinical adoption. One of the key concerns is the quality and diversity of data used to train AI models. AI systems are only as good as the data they are trained on and many current models are based on datasets that may not be representative of all patient populations. Pathology datasets need to include diverse demographic information to ensure AI tools perform well across different ethnic and age groups.
Another consideration is integrating pathology AI into existing clinical workflows. Radiology and pathology departments are fast-paced and highly regulated environments. Seamless integration with existing Picture Archiving and Communication Systems (PACS) and Laboratory Information Systems (LIS) is critical to ensure AI tools enhance rather than disrupt clinical practice. As AI gains traction in breast cancer screening, its integration into pathology workflows must be smooth and intuitive for clinicians.
Explainability of AI models is another challenge. While AI systems like those from Franklin.ai provide pathologists with insights into how findings are made, such as model confidence and patch-level localisation, full transparency in AI decision-making is still a work in progress. This kind of explainability helps ensure trust in AI’s findings and reinforces its role as a supporting tool for pathologists.
Looking Forward: Collaboration for AI in Pathology
The future of AI in breast cancer screening and diagnosis is bright. But overcoming the challenges of data diversity, workflow integration, and model transparency requires collaboration between AI developers, pathologists, regulatory bodies and healthcare IT teams. By focusing on robust validation, clinical utility and data diversity, we can ensure AI tools are safe and effective for breast cancer detection and pathology.
AI can revolutionise breast cancer diagnostics by improving early detection, streamlining workflows and reducing the burden on healthcare providers. As research and clinical trials continue to demonstrate AI’s capabilities, it’s essential to address the existing challenges so the healthcare system can fully utilise AI to enhance breast cancer diagnostics and pathology.
References and Further Reading
- World Health Organisation. "Breast cancer cases and deaths are projected to rise globally."https://www.iarc.who.int/wp-content/uploads/2025/02/pr361_E.pdfHomepage – IARC
- Breast Cancer Network Australia. "Breast cancer statistics in Australia."https://www.bcna.org.au/resource-hub/articles/breast-cancer-in-australia/BCNA Homepage
- The Guardian. "NHS to launch world's biggest trial of AI breast cancer diagnosis."https://www.theguardian.com/society/2025/feb/04/nhs-to-launch-worlds-biggest-trial-of-ai-breast-cancer-diagnosis
- MIT News. "AI model identifies certain breast tumour stages likely to progress to invasive cancer."https://news.mit.edu/2024/ai-model-identifies-certain-breast-tumor-stages-0722
- BreastCancer.org. "Using AI (Artificial Intelligence) to Detect Breast Cancer."https://www.breastcancer.org/screening-testing/artificial-intelligence
- BCRF. "Can AI and Machine Learning Revolutionise the Mammogram?"https://www.bcrf.org/blog/ai-breast-cancer-detection-screening/
