Brain MR Image Restoration using Hybrid LSTM-Unet Approaches

In medical imaging, magnetic resonance imaging (MRI) is a widely used non-invasive diagnostic tool. Accurate diagnosis is made more difficult by noise and abnormalities in MRI scans, which can severely reduce image quality. This research efficiently restores noisy brain MRI images by integrating U-Net and Convolutional Long Short-Term Memory (ConvLSTM) architectures in a revolutionary deep learning-based method. The ConvLSTM-UNet hybrid model improves picture quality by utilizing the spatial hierarchical structure of U-Net and the temporal coherence of MRI sequences. Based on our tests on different parameters like PSNR, SSIM and MSE, we have shown that the suggested approach improves the reliability of medical diagnoses by outperforming traditional methods in restoring high-quality brain MRI pictures.