Javascript must be enabled to continue!
Explainable Deep Learning for Breast Lesion Classification in Digital and Contrast-Enhanced Mammography
View through CrossRef
Background: Artificial intelligence (AI) emerges as a powerful tool to assist breast cancer screening; however, its integration into different mammographic modalities remains insufficiently explored. Digital Mammography (DM) is widely accessible but presents limitations in dense breast tissue, whereas Contrast-Enhanced Spectral Mammography (CESM) provides functional information that enhances lesion visualization. Understanding how deep learning models behave across these modalities, and determining whether their decision-making patterns remain consistent, is essential for equitable clinical adoption. Methods: This study evaluates three convolutional neural network (CNN) architectures, ResNet-18, DenseNet-121, and EfficientNet-B0, for binary classification of breast lesions using DM and CESM images from the public CDD-CESM dataset (2006 images, three diagnostic classes). The models are trained separately on DM and CESM using three classification tasks: Normal vs. Benign, Benign vs. Malignant, and Normal vs. Malignant. A 3-fold cross-validation scheme and an independent test set are employed. Training uses transfer learning with ImageNet weights, weighted binary cross-entropy (BCE) loss, and SHapley Additive exPlanations (SHAP) analysis to visualize pixel-level relevance of model decisions. Results: CESM yields higher performance in the Normal vs. Benign and Benign vs. Malignant tasks, whereas DM achieves the highest discriminative ability in the Normal vs. Malignant comparison (EfficientNet-B0: AUC = 97%, Accuracy = 93.15%), surpassing the corresponding CESM results (AUC = 93%, Accuracy = 85.66%). SHAP attribution maps reveal anatomically coherent decision patterns in both modalities, with CESM producing sharper and more localized relevance regions due to contrast uptake, while DM exhibits broader yet spatially aligned attention. Across architectures, EfficientNet-B0 demonstrates the most stable performance and interpretability. Conclusions: CESM enhances subtle lesion discrimination through functional contrast, whereas DM, despite its simpler acquisition and wider availability, provides highly accurate and explainable outcomes when combined with modern CNNs. The consistent SHAP-based relevance observed across modalities indicates that both preserve clinically meaningful information. To the best of our knowledge, this study is the first to directly compare DM and CESM under identical preprocessing, training, and evaluation conditions using explainable deep learning models.
Title: Explainable Deep Learning for Breast Lesion Classification in Digital and Contrast-Enhanced Mammography
Description:
Background: Artificial intelligence (AI) emerges as a powerful tool to assist breast cancer screening; however, its integration into different mammographic modalities remains insufficiently explored.
Digital Mammography (DM) is widely accessible but presents limitations in dense breast tissue, whereas Contrast-Enhanced Spectral Mammography (CESM) provides functional information that enhances lesion visualization.
Understanding how deep learning models behave across these modalities, and determining whether their decision-making patterns remain consistent, is essential for equitable clinical adoption.
Methods: This study evaluates three convolutional neural network (CNN) architectures, ResNet-18, DenseNet-121, and EfficientNet-B0, for binary classification of breast lesions using DM and CESM images from the public CDD-CESM dataset (2006 images, three diagnostic classes).
The models are trained separately on DM and CESM using three classification tasks: Normal vs.
Benign, Benign vs.
Malignant, and Normal vs.
Malignant.
A 3-fold cross-validation scheme and an independent test set are employed.
Training uses transfer learning with ImageNet weights, weighted binary cross-entropy (BCE) loss, and SHapley Additive exPlanations (SHAP) analysis to visualize pixel-level relevance of model decisions.
Results: CESM yields higher performance in the Normal vs.
Benign and Benign vs.
Malignant tasks, whereas DM achieves the highest discriminative ability in the Normal vs.
Malignant comparison (EfficientNet-B0: AUC = 97%, Accuracy = 93.
15%), surpassing the corresponding CESM results (AUC = 93%, Accuracy = 85.
66%).
SHAP attribution maps reveal anatomically coherent decision patterns in both modalities, with CESM producing sharper and more localized relevance regions due to contrast uptake, while DM exhibits broader yet spatially aligned attention.
Across architectures, EfficientNet-B0 demonstrates the most stable performance and interpretability.
Conclusions: CESM enhances subtle lesion discrimination through functional contrast, whereas DM, despite its simpler acquisition and wider availability, provides highly accurate and explainable outcomes when combined with modern CNNs.
The consistent SHAP-based relevance observed across modalities indicates that both preserve clinically meaningful information.
To the best of our knowledge, this study is the first to directly compare DM and CESM under identical preprocessing, training, and evaluation conditions using explainable deep learning models.
Related Results
Coexisting Granulomatous Mastitis and Breast Cancer: A Systematic Review
Coexisting Granulomatous Mastitis and Breast Cancer: A Systematic Review
Abstract
Introduction: Granulomatous mastitis (GM) is a rare inflammatory breast disease that mimics carcinoma. GM can coexist with breast cancer (BC), though the relationship rema...
Breast Carcinoma within Fibroadenoma: A Systematic Review
Breast Carcinoma within Fibroadenoma: A Systematic Review
Abstract
Introduction
Fibroadenoma is the most common benign breast lesion; however, it carries a potential risk of malignant transformation. This systematic review provides an ove...
Desmoid-Type Fibromatosis of The Breast: A Case Series
Desmoid-Type Fibromatosis of The Breast: A Case Series
Abstract
IntroductionDesmoid-type fibromatosis (DTF), also called aggressive fibromatosis, is a rare, benign, locally aggressive condition. Mammary DTF originates from fibroblasts ...
Hydatid Disease of The Brain Parenchyma: A Systematic Review
Hydatid Disease of The Brain Parenchyma: A Systematic Review
Abstarct
Introduction
Isolated brain hydatid disease (BHD) is an extremely rare form of echinococcosis. A prompt and timely diagnosis is a crucial step in disease management. This ...
Abstract P5-01-02: Glandular dose in contrast-enhanced dual-energy mammography
Abstract P5-01-02: Glandular dose in contrast-enhanced dual-energy mammography
Abstract
INTRODUCTION
Estimation of the Average Glandular Dose (AGD) absorbed by the breast during x-ray based examination is an established part of q...
Staging of breast cancer and the advanced applications of digital mammogram: what the physician needs to know?
Staging of breast cancer and the advanced applications of digital mammogram: what the physician needs to know?
Objective:
To study the role of advanced applications of digital mammogram, whether contrast-enhanced spectral mammography (CESM) or digital breast tomosynthesis ...
Role of Tomosynthesis with correlation to 2D Mammography in characterization of breast lesions in dense breast
Role of Tomosynthesis with correlation to 2D Mammography in characterization of breast lesions in dense breast
Abstract
Background
Digital Breast Tomosythesis is a new technology of digital mammography that enables the acquisition of three...
Abstract IA15: Neighborhood factors and changes in mammography screening rates in response to promotora intervention in Latinas
Abstract IA15: Neighborhood factors and changes in mammography screening rates in response to promotora intervention in Latinas
Abstract
Latinas have relatively low rates of mammography screening participation, which may contribute to later stage of breast cancer at diagnosis for Latinas, and...

