BoostCNN: Deep Learning AdaBoost-based Method for Easy and Difficult Nodule Classification in Ultrasound Images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Breast ultrasonography complements mammography but faces challenges with high false positive rates.
• Lack of consensus exists on ultrasound's primary role in breast cancer screening due to over-diagnosis and unnecessary treatments.
• False positives in breast ultrasound lead to significant patient distress and increased healthcare costs.

Purpose of the Study:

• To develop a novel AdaBoost-based ensemble method, BoostCNN, for improved breast cancer classification in ultrasound images.
• To reduce false positives and false negatives in breast cancer detection using ultrasound.
• To enhance the diagnostic accuracy and specificity of automated breast ultrasound analysis.

Main Methods:

• Implemented and evaluated a novel AdaBoost-based ensemble deep learning model named BoostCNN.
• Trained and tested BoostCNN and four other deep learning models on the BUS-BRA dataset.
• Validated the methodologies on external BUSI and BrEaSt datasets, assessing performance with varying ensemble sizes.

Main Results:

• The BoostCNN ensemble with nine models achieved 88.38% accuracy, 82.90% sensitivity, and 91.00% specificity on the BUS-BRA dataset.
• Combined ultrasonographer assessment with BoostCNN prediction yielded high performance, with 88.21% accuracy and 90.13% specificity.
• The proposed method demonstrated potential in enhancing specificity for breast cancer ultrasound classification.

Conclusions:

• BoostCNN offers a promising tool to improve the accuracy of breast cancer detection in ultrasound imaging.
• The ensemble method effectively addresses the challenge of false positives in breast ultrasound, potentially reducing unnecessary interventions.
• Further integration of AI tools like BoostCNN can support specialists in breast cancer diagnosis, improving patient outcomes.