This project aimed to revolutionize radiology workflows by developing RadioGen, a deep learning-based model that automates the generation of medical reports from chest X-ray images. The model combines cutting-edge image processing and natural language processing (NLP) techniques to produce accurate, standardized, and contextually rich radiology reports.
RadioGen leverages a multi-faceted architecture:
Multi-CNN Feature Extraction: Utilizes ResNet, EfficientNet, and MobileNet to capture diverse anatomical details, ensuring robust feature extraction for various conditions. NLP Integration: Employs RNN and Transformer models for coherent report generation, incorporating advanced medical terminology and context. Ensemble Learning: Combines predictions from multiple architectures, improving overall performance and reducing error rates. Key Achievements:
Achieved 88% accuracy, 89% precision, and 88% recall on chest X-ray classification tasks, identifying conditions like Normal, Pneumonia, COVID-19, and Tuberculosis. Automated medical report generation, reducing radiologists' workload while maintaining high clinical relevance. Ensured consistency and reduced variability in reporting through standardized language and structure.
https://www.kaggle.com/datasets/jtiptj/chest-xray-pneumoniacovid19tuberculosis