YOLO algorithm improves diagnostic performance of mammography: More than eyes

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Breast cancer (BC) is the most prevalent malignancy in women globally.
• Early detection and accurate diagnosis are critical for enhancing patient survival rates.

Purpose of the Study:

• To develop an integrated computer-aided diagnosis (CAD) system for automated lesion detection, segmentation, and classification in mammographic images.
• To enhance the diagnostic process for breast cancer (BC).

Main Methods:

• Utilized YOLOv5 as the object-detection model.
• Employed the Curated Breast Imaging Subset of the Digital Database for Screening Mammography (CBIS-DDSM) dataset.
• Applied data augmentation techniques and split data into training, validation, and testing sets (7:2:1 ratio).

Main Results:

• Achieved a mean average precision at 0.5 confidence threshold (mAP@0.5) of 0.417 and an F1-score of 0.46 for lesion detection.
• Obtained an area under the receiver operating characteristic (ROC) curve (AUC) of 0.90 for distinguishing benign from malignant lesions.
• The system processed images at a rate of 65 frames per second (fps).

Conclusions:

• The integrated CAD system demonstrates a combination of rapid processing and high accuracy in breast cancer lesion detection and classification.
• The system holds significant potential for clinical application in improving breast cancer diagnosis.