Image Denoising Using Quantum Deep Convolutional Generative Adversarial Network for Medical Images

A significant role is played by medical images in diagnosing diseases and planning the course of treatment. Noise can potentially degrade the quality of images which can lead to misdiagnosis. One of the oldest challenges in computer vision for restoring images that have been corrupted is image denoising. Generative adversarial networks (GANs) are among the most extensively used deep learning methods for various computer vision tasks. Utilizing an innovative quantum adversarial denoising architecture, denoised image samples are produced from a noisy distribution. In this paper, the authors employ an architecture of quantum deep convolutional generative adversarial networks (QDCGAN) for denoising medical images. The architecture of the DCGAN (deep convolutional generative adversarial networks) is augmented with a quantum computing layer to enhance the performance through quantum-generated inputs. The research is performed on the BraTS dataset via the TensorFlow Quantum platform. The study demonstrates that QDCGAN outperforms traditional methods. The proposed method achieves a better PSNR (peak signal-to-noise ratio) and SSIM (structural similarity index measure) value. The study underscores its effectiveness in improving the diagnostic quality of medical images with an 3.4% enhancement in SSIM and 7.35% in PSNR over existing methods, thereby offering tangible benefits for healthcare practitioners and patients alike.