Abstract - The use of medical image analysis is now an essential aspect of today’s healthcare systems due to early disease detection, accurate diagnosis, and sound clinical decision making. Nevertheless, current deep learning methods exhibit low interpretability, low generalization across different types of medical imaging datasets, and inability to perform in real time. Moreover, the lack of explainability in several convolutional and transformer models makes clinicians skeptical about their usage in critical healthcare settings. Therefore, this research offers a new Explainable Hybrid Vision Transformer with Generative Intelligence Network (EHVT-GIN) for real-time medical image analysis and diagnosis of diseases in an interpretable manner. The presented architecture combines Vision Transformer based global features extraction with local features learning using lightweight convolutional networks to maintain contextual and spatial information. The Generative AI module provides realistic augmentation of medical data samples to solve the class imbalance issue and increase feature diversity during training. The explainability component uses attention visualization and gradient attribution to provide clinically relevant diagnostic information, and the adaptive confidence calibration allows for more reliable predictions for multi-class diseases classification. Experiments on a number of public medical image datasets show that the proposed EHVT-GIN architecture surpasses other EfficientNet, Vision Transformer, and Hybrid CNN-Transformer architectures in terms of classification accuracy, which is achieved at 98.72% accuracy, 98.46% precision, 98.31% recall, 98.38% F1 score, and an average inference time of 17.1 ms per image. These findings prove that the proposed approach provides significant improvement in diagnostics, computation, robustness, and explainability, thus making it an appropriate choice for trustworthy real-time medical image analysis and disease diagnostics.