Deep Learning-Based Automated Reports for Breast Density Assessment in Mammography Images.

A novel hierarchical transfer learning to enhance breast cancer detection on screening mammography using deep learning.

Comparative Heatmap Explanations for Deep Learning-Based Breast Density Classification

Breast density is associated with a higher risk of developing breast cancer and complicates mammographic interpretation because dense tissue can obscure suspicious findings. Deep learning (DL) models are increasingly used for automated breast density classification, yet their limited interpretability remains a concern in safety-critical medical applications. In this paper, we compare several DL backbones for four-class BI-RADS density classification and analyze their predictions using three heatmap explanation methods—GradCAM, GradCAM++, and ScoreCAM. Our results show that transformer-based models achieve marginally stronger classification performance than the evaluated convolutional neural network (CNN) baselines, with the ViT model obtaining the best overall results and showing improved recognition of extremely dense cases. Qualitative heatmap analysis suggests that ScoreCAM produces the most spatially coherent and clinically plausible explanations, particularly for transformer-based models for which gradient-based maps are often diffuse or unstable. We further present an exploratory procedure that converts ScoreCAM heatmaps into coarse binary masks through thresholding, illustrating the potential of explanation maps for weak localization of dense tissue. Rather than claiming a validated segmentation framework, we position this step as a proof of concept that may support future studies on annotation-efficient dense-tissue localization.

DensiThAI: A Multi-View Deep Learning Framework for Breast Density Estimation Using Infrared Images.

Prediction Modeling for Breast Cancer Treatment Planning Using Contrast-Enhanced Mammography

It is suggested that integrating advanced CNN architectures with radiomics features can significantly enhance the accuracy and reliability of both lesion segmentation and treatment prediction in breast cancer, potentially leading to better clinical outcomes.

Hybrid Spatial and Frequency Feature Learning for Breast Cancer Analysis

Breast cancer is one of the leading causes of cancer-related mortality among women worldwide, where early diagnosis significantly improves treatment outcomes and survival rates. Full-field digital mammography (FFDM) is widely employed for breast cancer screening; however, accurate classification of mammographic abnormalities remains challenging due to tissue density variations, low-contrast regions, and subtle microcalcification patterns. Existing deep learning approaches often face limitations in effectively capturing both structural and texture-based information required for reliable lesion classification. To address these challenges, this paper proposes a hybrid spatial and frequency feature learning framework for breast cancer analysis using mammographic images. Initially, preprocessing techniques including normalization, noise reduction, contrast enhancement, and data augmentation are applied to improve image quality and robustness. A convolutional neural network-based multiscale feature extraction module is employed to capture spatial structural information from mammograms. Simultaneously, frequency-domain representations are extracted using adaptive wavelet-based decomposition to enhance fine-grained texture and microcalcification characteristics. Furthermore, an attention-guided feature fusion mechanism is introduced to integrate complementary spatial and frequency features for improved lesion representation and classification accuracy. Finally, the fused features are processed through a fully connected classification layer to distinguish benign and malignant breast lesions. Experimental evaluation conducted on the CBIS-DDSM dataset demonstrates that the proposed framework achieves superior performance with an accuracy of 99.25%, sensitivity of 99.58%, specificity of 98.88%, F1-score of 99.57%, and AUC of 99.15%. The obtained results demonstrate the effectiveness of the proposed hybrid learning framework in improving breast cancer classification and supporting computer-aided diagnosis systems for clinical applications.

Automated Breast Cancer Detection Using GLCM Texture Analysis and Ensemble Learning Techniques

Mammography is sometimes subject to intrinsic low contrast and noise, which results in a high degree of inter-observer variability among radiologists. This study investigates a classification system that focuses on textural feature analysis in the detection of malignancies at an early stage. Second-order statistical features, such as contrast, correlation, energy, and homogeneity, were used to detect the relationship between pixel intensities across space and reveal abnormal tissue growth using the Gray-Level Co-occurrence Matrix (GLCM). This study used 306 clinical mammograms (133 benign and 173 malignant) from the Curated Breast Imaging Subset of DDSM (CBIS-DDSM) to systematically benchmark five predictive models. Simpler models, such as Logistic Regression and SVM, demonstrated a low ability to determine a complex boundary in textual features, while XGBoost exhibited a higher discriminative ability. The XGBoost model achieved a maximum accuracy of 93.55% and a recall (sensitivity) rate of 94.29%. The findings demonstrate that integrating hand-crafted textural features and boosting techniques can mitigate false negative risks, offering a highly efficient and comprehensible instrument for clinical decision support in the screening of breast cancer.