BrainRotViT: Transformer-ResNet Hybrid for Explainable Modeling of Brain Aging from 3D sMRI

The paper presents a two-stage wavelet-driven masked autoencoder (WISE-MAE) framework designed for histopathology representation learning. It addresses the challenges of self-supervised learning in digital pathology by improving patch selection through a wavelet-informed strategy. This method enhances the model's ability to capture relevant tissue patterns, thereby aligning more closely with the diagnostic processes of pathologists.

This study presents a comprehensive benchmarking of convolutional neural networks (CNNs), vision transformers, and state-space mamba architectures for automated dental caries segmentation using panoramic radiographs. The research, utilizing the DC1000 dataset, reveals that the CNN-based DoubleU-Net outperformed other architectures, achieving the highest dice coefficient, mIoU, and precision, highlighting the effectiveness of simpler models in this domain.

The paper presents a hybrid deep learning framework for weed detection in precision agriculture, combining Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), and Graph Neural Networks (GNNs). This approach enhances robustness across various field conditions and employs a Generative Adversarial Network (GAN) for data augmentation, achieving an impressive accuracy of 99.33% on benchmark datasets. The model's architecture supports comprehensive feature representation, crucial for sustainable crop management.

The RS-CA-HSICT framework introduces a hybrid deep learning architecture that combines CNN and Transformer models to improve monkeypox detection. This innovative approach includes an HSICT block, a residual CNN module, and a spatial CNN block, enhancing feature extraction and long-range dependencies. The framework aims to provide detailed lesion information and reduce noise, addressing the challenges in accurately detecting monkeypox.

Dynamic Curriculum Learning for Spatiotemporal Encoding (DCL-SE) is introduced as an innovative framework aimed at enhancing high-dimensional neuroimaging analyses for clinical diagnosis. By utilizing Approximate Rank Pooling (ARP), DCL-SE efficiently encodes three-dimensional brain data into two-dimensional representations. The framework employs a dynamic curriculum learning strategy to progressively refine feature extraction, demonstrating effectiveness across various datasets, including Alzheimer's disease and brain tumor classification.

Bladder cancer, with a recurrence rate of up to 78%, poses significant challenges for post-operative monitoring. Traditional multi-sequence contrast-enhanced MRI scans are often difficult to interpret due to changes from surgery. This study introduces H-CNN-ViT, a new AI model designed to enhance bladder cancer recurrence prediction by utilizing a curated multi-sequence MRI dataset, which aims to improve diagnostic accuracy and patient management.

Deep neural networks are susceptible to adversarial perturbations that can lead to incorrect predictions. The paper introduces DeepDefense, a defense framework utilizing Gradient-Feature Alignment (GFA) regularization across multiple layers to mitigate this vulnerability. By aligning input gradients with internal feature representations, DeepDefense creates a smoother loss landscape, reducing sensitivity to adversarial noise. The method shows significant robustness improvements against various attacks, particularly on the CIFAR-10 dataset.

This study presents a deep learning framework for classifying functional groups in protein sequences, utilizing data from the Protein Data Bank (PDB). Four architectures were tested: Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory (BiLSTM), a CNN-BiLSTM hybrid, and CNN with Attention. The CNN model achieved the highest validation accuracy of 91.8%, highlighting its effectiveness in detecting localized motifs. Explainable AI techniques, such as Grad-CAM and Integrated Gradients, were employed to interpret predictions and identify biologically relevant sequence motifs.