Dual Generators and Dynamically Fused Discriminators Adversarial Network to Create Synthetic Coronary Optical Coherence Tomography Images for Coronary Artery Disease Classification

• DOI: 10.20944/preprints202503.0487.v1
• Publication Date: 2025-03-10T01:40:07Z
• Authors: Junaid Zafar; Faisal Sharif; Haroon Zafar

Deep neural networks have led to a substantial incursion for multifaceted classification tasks by making use of large-scale and diverse annotated datasets. However, optical coherence tomography (OCT) datasets in the cardiovascular imaging remains an uphill task. This research focuses on improving diversity generalization ability augmentation architectures while maintaining baseline accuracy coronary atrial plaques using novel dual generators dynamically fused discriminators conditional generative adversarial network (DGDFGAN). Our method is demonstrated augmented OCT dataset 6900 images. With generators, our provides outputs same input condition as each generator acts regularize other. In model, this mutual regularization enhances both generalize better across different features. The fusion one discriminator purposes hence avoiding need separate deep architecture. A loss functional including SSIM FID scores confirm that perfect synthetic image aliases are created. We optimize model via Grey Wolf optimizer during training. Furthermore, inter-comparison recorded SSID 0.9542±0.008 score 7 suggestive generation characteristics outperforms performance leading GANs architectures. trust approach practically viable thus assist professionals informed decision clinical settings.