Multi-Tier Kernel for Disease Prediction using Texture Analysis with MR Images

Denoising magnetic resonance images are traditionally done individually, introducing undesired artefacts like blurring. To solve this issue, this paper offers a unique adaptive interpolation architecture that simultaneously allows for image data augmentation, noise removal, and detail augmentation. The multi-tier kernel (MTK) algorithm adjusts the extrapolation weights based on mathematical features in magnetic resonance (MR) data. The MTK weight matrix is then adaptively sharpened, and a Rician bias corrective is used to reduce Rician noise and improve small details. After processing, the noise eliminates the bias produced by the asymmetric sources. The adaptive MTK, in this way, extends the zero ordering estimating methodology to higher regression power facilitating edge maintenance and detail restoration. Investigation outcomes using genuine and MR images (with/without noise) evidenced that the algorithm delivered good restoration outcomes than conventional deep-learning-based approaches but with fewer requirements and calculation burden.