AI Algorithm Detects One-Third of Missed Breast Cancers in 3D Mammography Screenings

AI Algorithm Shows Promise in Detecting Missed Breast Cancers

A groundbreaking study published in the journal Radiology has revealed that an artificial intelligence (AI) algorithm can significantly enhance the performance of digital breast tomosynthesis (DBT), also known as 3D mammography, in breast cancer screening. The research, conducted by Dr. Manisha Bahl and her team at Massachusetts General Hospital and Harvard Medical School, demonstrates that AI has the potential to reduce interval cancers by up to one-third 1

1

Understanding Interval Cancers and DBT

Interval breast cancers, which are symptomatic cancers diagnosed between regular screening mammography exams, often have poorer outcomes due to their aggressive nature and rapid growth. Digital breast tomosynthesis, a relatively new advanced screening technology, improves the visualization of breast lesions and can reveal cancers that may be obscured by dense tissue 2

2

Source: Medical Xpress

Study Findings and Methodology

The research team analyzed 1,376 cases, including 224 interval cancers in women who had undergone DBT screening. The AI algorithm, Lunit INSIGHT DBT v1.1.0.0, correctly localized 32.6% (73/224) of cancers that were previously undetected by radiologists 1

1

Dr. Bahl emphasized the significance of these findings, stating, "My team and I were surprised to find that nearly one-third of interval cancers were detected and correctly localized by the AI algorithm on screening mammograms that had been interpreted as negative by radiologists, highlighting AI's potential as a valuable second reader" 2

2

AI Performance and Accuracy

To ensure accurate assessment of the AI algorithm's performance, the researchers employed a lesion-specific analysis. This approach credits the AI only when it correctly identifies and localizes the exact site of the cancer, providing a more accurate reflection of its clinical performance 1

1

In a subset of 1,000 patients, including those with true-positive cancers and those with true-negative and false-positive outcomes, the algorithm demonstrated impressive accuracy:

• Correctly localized 84.4% of 334 true-positive cancers
• Correctly categorized 85.9% of 333 true-negative cases
• Correctly identified 73.2% of 333 false-positive cases as negative 2

  2

Related Stories

AI Shows Promise in Early Detection of Interval Breast Cancers, UCLA Study Finds

AI Boosts Breast Cancer Detection in Large-Scale German Study

AI Revolutionizes Breast Cancer Detection: New Tools Predict Risk and Catch Missed Tumors

AI mammography screening cuts aggressive breast cancer diagnoses by 12% in landmark trial

Implications for Cancer Detection and Treatment

Source: News-Medical

Interestingly, the study found that cancers detected by the AI algorithm tended to be larger and more likely to be lymph node-positive. Dr. Bahl noted, "These findings suggest that AI may preferentially detect more aggressive or rapidly growing tumors, or that it identifies missed cancers that were already advanced at the time of screening" 1

1

Future Prospects and Challenges

While the study's results are promising, Dr. Bahl cautioned that the real-world impact of AI in breast cancer screening will depend on radiologist adoption and validation across diverse clinical environments. The integration of AI into DBT screening workflows has the potential to enhance cancer detection and improve screening outcomes, but further research and implementation strategies will be necessary 2

2