Breast cancer pathology image recognition based on convolutional neural network.

Area of Science:

• Medical Imaging
• Computational Pathology
• Artificial Intelligence in Oncology

Background:

• Traditional pathological tissue analysis is time-consuming, labor-intensive, and prone to diagnostic errors.
• Accurate and efficient classification of breast cancer pathology images is crucial for timely diagnosis and treatment.

Purpose of the Study:

• To develop and evaluate a novel convolutional neural network (CNN)-based method for classifying and recognizing breast cancer pathology images.
• To overcome the limitations of conventional diagnostic methods using advanced deep learning techniques.

Main Methods:

• Employed ensemble learning by dividing images into multiple parts for data augmentation.
• Utilized the Inception-ResNet V2 model with transfer learning for feature extraction.
• Constructed a three-layer fully connected neural network for classification, integrating sub-image recognition results.

Main Results:

• The proposed CNN method achieved high accuracy rates of 99.75%, 98.31%, 98.51%, and 96.69% across four magnifications on the BreaKHis dataset.
• Outperformed established models including Inception-ResNet V2, ResNet101, DenseNet169, MobileNetV3, and EfficientNetV2.
• Demonstrated superior performance in classifying benign and malignant breast cancer lesions.

Conclusions:

• The CNN-based approach offers a significant advancement in automated breast cancer pathology image analysis.
• This method provides a more accurate, efficient, and reliable alternative to traditional diagnostic techniques.
• The study highlights the potential of deep learning in improving breast cancer diagnosis and patient outcomes.