Objective: The aim of this study was to evaluate the accuracy of machine and deep learning approaches on radiomics features obtained by Dynamic Contrast Enhanced Magnetic Resonance Imaging (DCE-MRI) and contrast enhanced mammography (CEM) in the characterization of breast cancer and in the prediction of the tumor molecular profile. Methods: A total of 153 patients with malignant and benign lesions were analyzed and underwent MRI examinations. Considering the histological findings as the ground truth, three different types of findings were used in the analysis: (1) benign versus malignant lesions; (2) G1 + G2 vs. G3 classification; (3) the presence of human epidermal growth factor receptor 2 (HER2+ vs. HER2−). Radiomic features (n = 851) were extracted from manually segmented regions of interest using the PyRadiomics platform, following IBSI-compliant protocols. Highly correlated features were excluded, and the remaining features were standardized using z-score normalization. A feature selection process based on Elastic Net regularization (α = 0.5) was used to reduce dimensionality. Synthetic balancing of the training data was applied using the ROSE method to address class imbalance. Model performance was evaluated using repeated 10-fold cross-validation and AUC-based metrics. Results: Among the 153 patients enrolled in the studies, 113 were malignant lesions. Among the 113 malignant lesions, 32 had high grading (G3) and 66 had the HER2+ receptor. Radiomic features derived from both CEM and DCE-MRI showed strong discriminative performance for malignancy detection, with several features achieving AUCs above 0.80. Gradient Boosting Machine (GBM) achieved the highest accuracy (0.911) and AUC (0.907) in differentiating benign from malignant lesions. For tumor grading, the neural network model attained the best accuracy (0.848), while LASSO yielded the highest sensitivity (0.667) for detecting high-grade tumors. In predicting HER2+ status, the neural network also performed best (AUC = 0.669), with a sensitivity of 0.842. Conclusions: Radiomics-based machine learning models applied to multiparametric CEM and DCE-MRI images offer promising, non-invasive tools for breast cancer characterization. The models effectively distinguished benign from malignant lesions and showed potential in predicting histological grade and HER2 status. These results demonstrate that radiomic features extracted from CEM and DCE-MRI, when analyzed through machine and deep learning models, can support accurate breast cancer characterization. Such models may assist clinicians in early diagnosis, histological grading, and biomarker assessment, potentially enhancing personalized treatment planning and non-invasive decision-making in routine practice.

In the era of precision medicine, effective data visualisation plays a pivotal role in supporting clinical decision-making by translating complex, multidimensional datasets into intuitive and actionable insights. This paper explores the foundational principles of visual perception, with a specific focus on pre-attentive attributes such as colour, shape, size, orientation, and spatial position, which are processed automatically by the human visual system. Drawing from cognitive psychology and perceptual science, we demonstrate how these attributes can enhance the clarity and usability of medical visualisations, reducing cognitive load and improving interpretive speed in high-stakes clinical environments. Through detailed case studies and visual examples, particularly within the field of oncology, we highlight best practices and common pitfalls in the design of dashboards, nomograms, and interactive platforms. We further examine the integration of advanced tools—such as genomic heatmaps and temporal timelines—into multidisciplinary workflows to support personalised care. Our findings underscore that visually intelligent design is not merely an aesthetic concern but a critical factor in clinical safety, efficiency, and communication, advocating for user-centred and evidence-based approaches in the development of health data interfaces.

Deep-learning models applied to contrast-enhanced mammography (CEM) are known to be highly sensitive to the input image representation. However, preprocessing is often treated as a secondary step and rarely analyzed as an independent design variable. In this work, we present a systematic engineering analysis of a deterministic, label-independent preprocessing pipeline for CEM images. The pipeline integrates intensity normalization, global histogram matching, local contrast enhancement, denoising, and anatomically constrained breast masking. Using a controlled experimental design, identical deep-learning architectures were trained under different input representations to isolate the impact of preprocessing on classification performance and stability. Across convolutional neural network architectures, anatomically constrained preprocessing consistently improves discrimination performance, reduces variability across cross-validation folds, and enhances training stability. Breast mask-based representations demonstrate substantial gains in AUROC and AUPRC compared to raw DICOM inputs. These findings highlight image preprocessing as a first-class engineering component in medical AI pipelines. Breast masking significantly improves robustness and generalization, independently of network architecture complexity. From a clinical perspective, improving model robustness and sensitivity to malignant lesions may contribute to more reliable AI-assisted decision support in contrast-enhanced mammography, particularly in settings characterized by acquisition variability and heterogeneous patient populations.

Background: This study investigates the impact of anatomically constrained preprocessing and deep learning architecture selection on benign versus malignant breast lesion classification in contrast-enhanced mammography (CEM), with the goal of improving robustness and clinical reliability across heterogeneous data sources. Methods: In this retrospective multicenter study, CEM images from 300 patients (314 lesions) were combined with 1003 publicly available CEM images, yielding a total of 1120 breast cases. Automatic breast segmentation was performed using the LIBRA framework to generate breast-mask images. Eleven deep learning models, including classical convolutional neural networks, attention-based networks, hybrid convolutional neural networks (CNNs), Transformer architectures, and mammography-specific models, were trained and evaluated using both original DICOM images and breast-mask inputs. Performance was assessed using accuracy, balanced accuracy, sensitivity, specificity, AUROC, and AUPRC on cross-validation and independent test sets. Hyperparameter optimization was conducted for the best-performing architecture. Results: Models trained on breast-mask images consistently outperformed those trained on original DICOM images across all architectures and metrics, with AUROC improvements ranging from +0.06 to +0.21. Among all models, ResNet50 trained on breast-mask images achieved the best performance (AUROC = 0.931; AUPRC = 0.933; balanced accuracy = 0.834), further improved after optimization (balanced accuracy = 0.886; sensitivity = 0.842; specificity = 0.930). Classical CNN architectures demonstrated performance comparable to or exceeding that of more complex hybrid CNN–Transformer models when anatomically focused preprocessing and rigorous optimization were applied. Conclusions: Anatomically constrained preprocessing through breast-mask segmentation substantially enhances deep learning performance and stability in CEM-based breast lesion classification. These findings indicate that input representation quality and training optimization are critical determinants of clinically relevant performance, often outweighing architectural complexity, and may support more reliable AI-assisted decision support in CEM workflows.