AUTOMATED DETECTION OF CHRONIC DISEASES FROM MEDICAL IMAGES USING ARTIFICIAL INTELLIGENCE

Abstract

Recent developments in Artificial Intelligence and deep learning have brought a major transformation to the field of medical imaging allowing for automatic, accurate and large-scale disease identification. The present study introduces a diagnostic framework based on Convolution Neural Networks (CNNs) enhanced through transfer learning techniques using established models such as VGG16, ResNet50, and InceptionV3. The framework is designed for the automated recognition of chronic diseases from imaging modalities including X-rays, magnetic resonance imaging (MRI) and computed tomography (CT) scans. Experiments were carried out on publicly accessible datasets such as the NIH Chest X-ray repository and Kaggle Retinopathy dataset. The proposed model achieved an overall accuracy of 93.4%, F1-score of 0.91 and ROC–AUC value of 0.95 across multiple disease categories including diabetic retinopathy, lung cancer, and Alzheimer’s disease. The inclusion of rigorous pre-processing and diverse data augmentation strategies improved the model’s robustness and adaptability. Additionally, fine-tuning of pre-trained networks contributed to better generalization across varied medical images. The findings indicate that the developed AI-driven diagnostic framework performs comparably to experienced radiologists, while maintaining cost efficiency and scalability. These outcomes highlight the growing significance of AI in medical diagnostics and affirm its potential for integration into routine clinical practices to enhance early detection and patient care.