Local Synthesis of Healthy Brain Tissue Using an Enhanced 3D Pix2Pix Model for Medical Image Inpainting

The restoration of voided regions in three-dimensional (3D) brain Magnetic Resonance Imaging (MRI) data is a critical task in medical image processing. This paper proposes a 3D pix2pix GAN modeling strategy for 3D brain MRI image inpainting, with the goal of improving the quality and diagnostic utility of medical images. The proposed method renders advanced algorithms to accurately predict and reconstruct the voided regions of 3D brain MRI scans. The inpainting outcomes are assessed via online evaluation on BraTS-2023 inpainting challenge data using established metrics, such as the Mean Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR), and Structural Similarity Index (SSIM). The experimental results demonstrate the effectiveness of the method in achieving a mean SSIM, PSNR, and MSE score of 0.70, 17.52, and 0.049, respectively, indicating spatial coherence and visual accuracy. The improved PSNR and reduced MSE further underscore the quality enhancement achieved by the proposed inpainting method. This work contributes to the field of medical imaging by addressing the crucial challenge of 3D brain MRI image inpainting and demonstrates its potential to enhance medical image analysis, diagnosis, and patient care.