Search

Rare diseases identified via chest radiography—such as spontaneous pneumothorax, solitary pulmonary nodules, pleural effusions, and cardiomegaly—occur far less frequently than common conditions like pneumonia or chronic obstructive pulmonary disease. Deep learning models require large, balanced datasets for reliable performance, yet rare pathologies remain underrepresented in clinical repositories, limiting real-world deployability. Conventional augmentation methods (geometric and intensity transformations, elastic deformations) add limited variability without creating new pathological patterns. GAN-based approaches can generate synthetic images but often suffer from mode collapse and unrealistic artifacts that reduce lesion fidelity, restricting their effectiveness for rare disease augmentation. We propose a framework based on denoising diffusion probabilistic models (DDPMs) for conditional synthesis of high-fidelity chest X-ray images. The model supports generation conditioned on class labels, segmentation masks, or text prompts, enabling controlled synthesis of rare pathologies and improving dataset balance. The framework includes a forward diffusion process, a U-Net-based reverse denoising model with attention, a multi-modal conditioning mechanism, a lesion-preserving loss function, and an augmentation pipeline combining real and synthetic data. This allows control over lesion type, location, size, and severity, reducing class imbalance and improving classifier performance on rare diseases, as validated through AUC improvements and radiologist assessment. Overall, diffusion-based models provide a scalable and clinically relevant solution for rare disease augmentation in chest radiography, overcoming key limitations of traditional and GAN-based methods and enabling effective use of datasets such as CheXpert, MIMIC-CXR, and ChestX-ray14.

Chest X-ray report generation is time-consuming and contributes to radiologist workload and burnout, motivating the need for AI systems that can reduce cognitive burden while preserving clinical accuracy. Although encoder-decoder models can generate reports from images, they often suffer from hallucinations, producing findings that are not present or missing real abnormalities due to lack of explicit grounding in evidence, making them unreliable for clinical use. To address this, we propose a cross-modal retrieval framework that generates reports by retrieving and assembling clinically validated sentences from existing radiology reports rather than generating text from scratch. The system uses contrastive learning to align chest X-ray image patches with report sentences in a shared embedding space, enabling retrieval of the most relevant clinical descriptions. A patch encoder extracts visual features, a sentence encoder represents report text, and a retrieval module identifies semantically matching sentences, which are then composed into a coherent final report. Because all outputs are sourced from real clinical reports, the method substantially reduces hallucinations while improving factual reliability and interpretability. This retrieval-based approach offers a scalable and safer alternative to generative models and can be evaluated on datasets such as MIMIC-CXR and CheXpert for clinical accuracy and retrieval performance.