Здравоохранение: текущие обзоры

Абстрактный

Development and Analysis of Filtering Algorithm Applied to T1 Weighted Clinical Brain Magnetic Resonance Images

Vedant Shulka, Amanora Kandivali, Bhakti Shakti Sakinaka

Non-Linear filters are used to filter out the artifacts and noise present in MR data. The balance between signal preservation and noise reduction makes restoration of MR data a complex task. Application of non-linear filters such as median and Non-Local Filter (NLM) filter converts the right-skewed Rician Distribution into un-skewed Gaussian distribution. It is evident that NLM filter gives better results than Bilateral and Median filter. As the distribution is un-skewed after application of non-linear filters, standard linear filters such as Gaussian and Wiener filters were applied and results were drawn. A linear combination of NLM and Gaussian filter gives satisfactory results. The experimentation was performed on 40 clinical images and NLM filter was found to have the best results. The Image Quality Indices used for comparison are Peak Signal-to-Noise Ratio (PSNR), Root Mean Squared Error (RMSE), Structural Similarity Index (SSIM) and Entropy. The experimentation was performed on MATLAB 2019a.