Weakly Supervised Pneumonia Localization from Chest X-Rays Using Deep Neural Network and Grad-CAM Explanations

A recent study analyzed the multi-modal approach in the Vision Transformer (EVA-ViT) image encoder combined with LlaMA and ChatGPT large language models (LLMs) to address hallucination issues and enhance disease detection in chest X-ray images. The research utilized the NIH Chest X-ray dataset, comparing image-based and text-based retrieval-augmented generation (RAG) methods, revealing that text-based RAG effectively mitigates hallucinations while image-based RAG improves prediction confidence.

A new study has introduced an explainable two-stage deep learning framework for assessing pericoronitis in panoramic radiographs, utilizing YOLOv8 for anatomical localization and a modified ResNet-50 for pathological classification. The system achieved high precision and alignment with radiologists' diagnostic impressions, enhancing interpretability through Grad-CAM visualizations.

A novel deep learning framework named TIMM-ProRS has been introduced to enhance the prognosis and risk stratification of diabetic retinopathy (DR), a condition that threatens the vision of millions worldwide. This framework integrates Vision Transformer, Convolutional Neural Network, and Graph Neural Network technologies, utilizing both retinal images and temporal biomarkers to achieve a high accuracy rate of 97.8% across multiple datasets.

A new study has developed a pre-trained foundation model framework aimed at improving the classification of extramural vascular invasion (EVI) and mesorectal fascia invasion (MFI) in rectal cancer using multiplanar MRI scans. This framework was evaluated using 331 pre-treatment T2-weighted MRI scans from three European hospitals, addressing the limitations of subjective visual assessments and inter-institutional variability in diagnostics.