AI in Radiology: Improving Breast Cancer Rates

For millions of women, the anxiety of waiting for mammogram results is a familiar burden. This stress is often compounded by “false positives,” where a scan suggests cancer is present when it is not. New breakthroughs in artificial intelligence are now changing this dynamic. Recent clinical trials and FDA-cleared technologies are proving that AI can significantly reduce these false alarms while simultaneously catching invasive cancers that human eyes might miss.

The High Cost of False Positives

To understand why AI is such a vital addition to radiology, you must first look at the current limitations of screening. Standard digital mammography is effective, but it is not perfect.

Statistics indicate that if a woman undergoes annual mammograms for ten years, she has roughly a 50% chance of receiving at least one false positive recall. This results in a “callback.” This callback leads to additional imaging, unnecessary biopsies, high out-of-pocket costs, and significant psychological distress.

Radiologists face a difficult balancing act. If they are too aggressive in flagging potential tumors, false positives skyrocket. If they are too conservative, they might miss an aggressive cancer. This is where Artificial Intelligence steps in as a “second set of eyes.”

Evidence from the MASAI Trial

The most compelling evidence for AI in breast cancer screening comes from Sweden. In 2023, researchers at Lund University published the interim results of the MASAI (Mammography Screening with Artificial Intelligence) trial in The Lancet Oncology. This was the first randomized controlled trial to evaluate AI in a real-world national screening program.

The study involved over 80,000 women. The results were concrete and promising:

  • Increased Detection: The group screened with AI support saw a 20% increase in cancer detection compared to the group screened by two radiologists without AI.
  • Workload Reduction: The use of AI reduced the screen-reading workload for radiologists by 44%.
  • Safety: The rate of false positives remained consistent with standard double-reading, meaning the AI found more cancer without triggering a wave of unnecessary anxiety for healthy patients.

Lead researcher Dr. Kristina LĂĄng noted that the technology helps radiologists focus their attention where it is needed most, acting as an advanced triage system.

How the Technology Works

Unlike basic computer-aided detection (CAD) systems used in the early 2000s, modern AI uses Deep Learning. These algorithms are trained on millions of historical images where the diagnosis is known. They learn to recognize subtle patterns in breast tissue density that are invisible to the human eye.

Several companies are currently leading this charge with specific, FDA-cleared or CE-marked tools:

Kheiron Medical Technologies

Based in the UK, Kheiron developed a tool called Mia. In a study involving over 25,000 women in Hungary, Mia acted as a second independent reader. The system flagged potential malignancies that radiologists could review. It proved particularly effective at identifying small, invasive cancers that are often obscured in women with dense breast tissue.

Google Health and DeepMind

In a pivotal study published in Nature, Google Health developed an AI model that was tested on datasets from the UK and the USA. The results showed a tangible reduction in errors:

  • In the USA dataset, the AI system reduced false positives by 5.7%.
  • It also reduced false negatives (missed cancers) by 9.4%.

ScreenPoint Medical

This company produces Transpara, an AI system that provides a “decision support” score. It ranks mammograms on a scale of 1 to 10 based on the likelihood of cancer. A score of 10 alerts the radiologist to look extremely closely. Studies have shown that radiologists using Transpara match the performance of two independent radiologists reading the same scan.

Improving Accuracy for Dense Tissue

One of the greatest challenges in mammography is dense breast tissue. On a mammogram, fat appears dark, while glandular tissue and tumors both appear white. This creates a masking effect, often described as “looking for a snowball in a blizzard.”

AI ignores the visual camouflage. It analyzes pixel-level data to distinguish between benign dense tissue and malignant growth. For women with dense breasts (about 40% of women over age 40), AI integration effectively lowers the threshold for detecting cancer early, potentially reducing the need for more invasive MRI screenings solely for clarification.

The Future of the "Second Reader"

In the United States, the standard of care is usually one radiologist interpreting a scan. In Europe, the standard is often “double reading,” where two doctors review every image.

AI is poised to democratize the “double reading” standard. For clinics in the US, the AI acts as that second doctor. It flags regions of interest (ROI) on the screen. If the AI marks a spot that the radiologist initially scanned past, the doctor is prompted to look again.

This workflow does not remove the doctor from the equation. Instead, it prevents fatigue-related errors. Radiologists often view hundreds of images in a single sitting. Decision fatigue is real, but AI does not get tired. By handling the obvious negative cases (scans with no signs of cancer), AI frees up the radiologist’s mental energy to analyze the complex, ambiguous cases.

Frequently Asked Questions

Will AI replace my radiologist? No. AI is a support tool, not a replacement. A board-certified radiologist still reviews the images and makes the final diagnosis. The AI simply highlights suspicious areas to ensure nothing is missed.

Does insurance cover AI-enhanced mammograms? Currently, most insurance plans cover the standard mammogram fee. In many hospitals, the use of AI is becoming part of the standard operational overhead and is not billed separately to the patient. However, some specialized advanced analyses might carry an out-of-pocket cost depending on the facility.

Is this technology available now? Yes. FDA-cleared AI tools like iCAD’s ProFound AI and Hologic’s Genius AI are already deployed in thousands of imaging centers across the United States and Europe.

Does AI result in more biopsies? The goal of high-quality AI is to decrease unnecessary biopsies. By better distinguishing between benign cysts and malignant tumors, AI helps radiologists dismiss non-cancerous findings that might otherwise have looked suspicious.