How Artificial Intelligence is Revolutionizing Breast Cancer Detection

By Campion Quinn, MD

Breast cancer detection has entered a new era with the advent of artificial intelligence (AI). For decades, radiologists and pathologists have relied on human expertise to identify cancerous lesions in imaging and tissue samples. However, human interpretation is not infallible—fatigue, complexity, and variability can lead to errors. AI is transforming this landscape by improving diagnostic accuracy, reducing false positives, and streamlining workflows. In this essay, we’ll explore how AI works, highlight key products, and illustrate its practical benefits for clinicians in straightforward terms designed for physicians without a background in computer science.

How AI Works in Breast Cancer Detection

In breast cancer detection, AI employs machine learning, which learns patterns by analyzing vast datasets. Imagine teaching a resident to recognize malignant findings by showing them thousands of cases. Over time, they would become adept at identifying suspicious patterns. AI does the same, except it learns at a scale and speed unmatched by humans.

For instance, AI analyzes mammograms or pathology slides by comparing them to a database of images. It looks for telltale signs—such as irregular masses or unusual cellular arrangements—and flags them for further review. Think of AI as a diagnostic partner, consistently applying evidence-based criteria to each case.

AI in Mammography

Mammography is one of the primary tools for early breast cancer detection. However, interpreting these images is challenging, particularly in patients with dense breast tissue. AI products have emerged to assist radiologists:

1. Genius AI Detection (Hologic): This system reviews tomosynthesis (3D mammography) images slice by slice. This system meticulously analyzes each slice of tomosynthesis images to identify potential lesions. It has demonstrated a sensitivity of 94%, significantly aiding radiologists in detecting subtle cancers. Highlighting areas of concern helps radiologists focus on potential cancers without overlooking subtle abnormalities.

2. Transpara (ScreenPoint Medical): Transpara scores mammograms based on their likelihood of malignancy. Radiologists can use these scores to prioritize high-risk cases, like triaging emergency department patients.

3. Lunit INSIGHT MMG: This software identifies lesions that human readers may miss. Studies indicate that it can identify breast cancer with an accuracy comparable to radiologists, achieving a sensitivity of 96%.  It acts as a vigilant assistant, ensuring no detail escapes attention.

These tools reduce false positives by minimizing overdiagnosis of benign findings, sparing patients unnecessary biopsies and anxiety. At the same time, they improve sensitivity, catching cancers that might otherwise go undetected.

AI in Pathology

Pathologists play a crucial role in diagnosing breast cancer by examining tissue samples under the microscope. AI enhances this process by automating time-consuming tasks and improving diagnostic precision. Key products include:

1. Paige Breast Suite: This tool identifies cancerous areas in tissue slides and quantifies biomarkers such as HER2, ER, and PR. It provides pathologists with detailed analyses that support personalized treatment decisions.  It aims to reduce subjectivity and manual workload, thereby enhancing diagnostic confidence and efficiency.  

2. PathAI’s AIM-HER2: PathAI helps standardize the evaluation of HER2, a critical biomarker for guiding therapy. It reduces variability between pathologists and enhances the accuracy of biomarker assessment, enhancing the precision of pathology assessments and supporting personalized treatment strategies.

Think of these AI systems as "pathology concierges" that pre-screen slides and highlight areas of interest, freeing pathologists to focus on complex cases requiring nuanced judgment.

Improving Workflow for Clinicians

AI isn’t just a diagnostic tool—it’s a workflow enhancer. Radiologists, for example, often interpret hundreds of mammograms daily, a workload that can lead to fatigue and errors. AI improves workflow in several ways:

1. Triage: Tools like Transpara prioritize high-risk cases, ensuring the most concerning scans are reviewed promptly.

2. Consistency: Unlike humans, AI applies the same diagnostic criteria to every case, reducing variability.

3. Speed: AI reduces the time required for review by pre-analyzing images, allowing clinicians to focus on interpretation and decision-making.

Pathologists also benefit from AI’s efficiency. Once labor-intensive tasks such as counting mitotic figures or assessing receptor status can now be automated, improving speed and accuracy.

Addressing Health Disparities

AI can potentially reduce disparities in breast cancer care by bringing expert-level diagnostics to underserved regions. For example, facilities with limited access to experienced radiologists or pathologists can augment their diagnostic capabilities with AI. This democratization of expertise can improve outcomes for patients in rural or low-resource settings.

Ethical and Regulatory Considerations

While AI holds great promise, its implementation requires careful oversight. Poorly trained algorithms or biases in data can lead to inaccuracies, potentially exacerbating disparities. Regulatory bodies, such as the FDA, play a critical role in ensuring these tools are safe and effective for clinical use. Clinicians must also remain vigilant, treating AI as an adjunct rather than replacing human expertise.

Conclusion

AI is revolutionizing breast cancer detection by enhancing accuracy, reducing false positives, and streamlining workflows. Tools like Genius AI Detection, Transpara, and Paige Breast Suite exemplify how technology can complement the expertise of radiologists and pathologists. By addressing clinician fatigue, standardizing diagnostics, and expanding access to care, AI is transforming the detection of breast cancer and the broader practice of medicine.

As we embrace these advancements, we must remain mindful of their limitations. We must ensure that AI is a partner—not a substitute—in the fight against breast cancer.