Convolutional Feature Descriptor Selection for Mammogram Classification

Breast cancer was the most commonly diagnosed cancer among women worldwide in 2020. Recently, several deep learning-based classification approaches have been proposed to screen breast cancer in mammograms. However, most of these approaches require additional detection or segmentation annotations. Meanwhile, some other image-level label-based methods often pay insufficient attention to lesion areas, which are critical for diagnosis. This study designs a novel deep-learning method for automatically diagnosing breast cancer in mammography, which focuses on the local lesion areas and only utilizes image-level classification labels. In this study, we propose to select discriminative feature descriptors from feature maps instead of identifying lesion areas using precise annotations. And we design a novel adaptive convolutional feature descriptor selection (AFDS) structure based on the distribution of the deep activation map. Specifically, we adopt the triangle threshold strategy to calculate a specific threshold for guiding the activation map to determine which feature descriptors (local areas) are discriminative. Ablation experiments and visualization analysis indicate that the AFDS structure makes the model easier to learn the difference between malignant and benign/normal lesions. Furthermore, since the AFDS structure can be regarded as a highly efficient pooling structure, it can be easily plugged into most existing convolutional neural networks with negligible effort and time consumption. Experimental results on two publicly available INbreast and CBIS-DDSM datasets indicate that the proposed method performs satisfactorily compared with state-of-the-art methods.

Authors

• Dong Li University of Science and Technology of China
• Lei Zhang
• Jianwei Zhang College of Computer Science, Sichuan University, Chengdu 610000, China
• Xingyu Xie

Keywords

• Mammography
• Lesions
• Breast cancer
• Feature extraction
• Task analysis
• Annotations
• Image segmentation
• Descriptive Characteristics
• Mammogram Classification
• Convolutional Neural Network
• Local Area
• Feature Maps
• Lesion Area
• Activation Maps
• Breast Cancer Diagnosis
• Adaptive Selection
• Secret Sharing
• Precise Annotation
• Receiver Operating Characteristic Curve
• Malignant Lesions
• Object Detection
• Bounding Box
• Small Lesions
• Average Pooling
• Deep Features
• Large Lesions
• Malignant Samples
• Probability Of Malignancy
• Class Activation Maps
• Channel Values
• Screening Mammograms
• Ground Truth Segmentation
• Mean Threshold
• Fixed-length Vector
• Regression Layer
• Computer-aided Diagnosis
• Discriminative representation
• feature descriptor
• activation map