Denoising Magnetic Resonance Images with Rician Noise Based On Deep Learning
Priyanka Datta, Rajesh Rohilla · 2022 International Conference on Computational Intelligence and Sustainable Engineering Solutions (CISES) · 2022
Image quality, which deteriorates primarily due to noise and distortions, is critical for the effectiveness of magnetic resonance imaging (MRI) analysis. This distortion is caused because of poor imaging settings or transmission system interference. MRI scans are typically corrupted by Rician noise when acquisition takes place. This noise should be eliminated from MRI as a pre-processing phase. This paper proposes MRI denoising using deep neural network (DNN). To extract image features from noised image, the model employs a collection of convolutions. The model too uses an encoder-decoder arrangement to preserve important image features whereas rejecting irrelevant features. To reduce noise in MRI, this paper proposed DNN based technique and this technique was compared with various denoising techniques such as median filter, gaussian filter, average filter, nonlocal means filter, bilateral filter, wavelet smoothed filters. These approaches were compared using measures like peak signal to noise ratio (PSNR), Mean Square Error (MSE), Root Mean Square Error (RMSE). The results demonstrates that the suggested model can efficiently denoise MRI scans while preserving essential image features.