Leandro Alves Neves

Hybrid convolutional neural networks (CNNs) and Transformer-based architectures have demonstrated strong potential in histopathological image analysis by combining local feature extraction with global context modeling. This dual capability is critical in computational pathology, where accurate interpretation depends on both fine-grained morphological details and broad tissue-level patterns. To systematically assess how these hybrid models have been applied, this scoping review maps the current landscape of CNN-Transformer architectures across classification, segmentation, and detection tasks in histopathological images. Following the PRISMA-ScR guidelines and the Joanna Briggs Institute methodology, a comprehensive literature search was conducted across Web of Science, IEEE Xplore, Scopus, and PubMed. A total of 39 peer-reviewed journal articles published from 2020 onward met the inclusion criteria. To guide the analysis, a task-specific taxonomy was proposed to classify hybrid architectures based on their primary application, offering a structured framework for model comparison and evaluation. Thus, the review synthesizes key aspects of model implementation, including architectural designs, training strategies, dataset characteristics, and evaluation protocols. Findings reveal that hybrid models frequently outperform conventional approaches; however, critical limitations persist, particularly regarding external validation, model interpretability, and reproducibility. Notably, only a small subset of studies employed rigorous validation schemes or integrated explainability into clinical workflows. Moreover, by consolidating evidence and exposing methodological gaps, this review contributes a comprehensive foundation for advancing hybrid deep learning models in digital pathology. It offers actionable insights to inform the development of scalable, transparent, and clinically aligned computational tools for histopathological image analysis.

Deep learning models such as convolutional neural networks (CNNs) and vision transformers (ViTs) perform well in histological image classification, but often lack interpretability. We introduce a unified framework that adds an attention branch and CAM Fostering, an entropy-based regularizer, to improve Grad-CAM visualizations. Six backbone architectures (ResNet-50, DenseNet-201, EfficientNet-b0, ResNeXt-50, ConvNeXt, CoatNet-small) were trained, with and without our modifications, on five H&E-stained datasets. We measured explanation quality using coherence, complexity, confidence drop, and their harmonic mean (ADCC). Our method increased the ADCC in five of the six backbones; ResNet-50 saw the largest gain (+15.65%), and CoatNet-small achieved the highest overall score (+2.69%), peaking at 77.90% on the non-Hodgkin lymphoma set. The classification accuracy remained stable or improved in four models. These results show that combining attention and entropy produces clearer, more informative heatmaps without degrading performance. Our contributions include a modular architecture for both convolutional and hybrid models and a comprehensive, quantitative explainability evaluation suite.

Generative adversarial networks (GANs) create images by pitting a generator (G) against a discriminator (D) network, aiming to find a balance between the networks. However, achieving this balance is difficult because G is trained based on just one value representing D’s prediction, and only D can access image features. We introduce a novel approach for training GANs using explainable artificial intelligence (XAI) to enhance the quality and diversity of generated images in histopathological datasets. We leverage XAI to extract feature information from D and incorporate it into G via the loss function, a unique strategy not previously explored in this context. We demonstrate that this approach enriches the training with relevant information and promotes improved quality and more variability in the artificial images, decreasing the FID by up to 32.7% compared to traditional methods. In the data augmentation task, these images improve the classification accuracy of Transformer models by up to 3.81% compared to models without data augmentation and up to 3.01% compared to traditional GAN data augmentation. The Saliency method provides G with the most informative feature information. Overall, our work highlights the potential of XAI for enhancing GAN training and suggests avenues for further exploration in this field.

Domain Generation Algorithms (DGAs) are algorithms present in most malware used by botnets and advanced persistent threats. These algorithms dynamically generate domain names to maintain and obfuscate communication between the infected device and the attacker’s command and control server. Since DGAs are used by many threats, it is extremely important to classify a given DGA according to the threat it is related to. In addition, as new threats emerge daily, classifier models tend to become obsolete over time. Deep neural networks tend to lose their classification ability when retrained with a dataset that is significantly different from the initial one, a phenomenon known as catastrophic forgetting. This work presents a computational scheme composed of a deep learning model based on CNN and natural language processing and an incremental learning technique for class increment through transfer learning to classify 60 DGA families and include a new family to the classifier model, training the model incrementally using some examples from known families, avoiding catastrophic forgetting and maintaining metric levels. The proposed methodology achieved an average precision of 86.75%, an average recall of 83.06%, and an average F1 score of 83.78% with the full dataset, and suffered minimal losses when applying the class increment.