Advancement in Denoising MRI Images via 3D-GAN Model with Direction Coupled Magnitude Histogram Consistency Loss

Abstract

The diagnostics of medical pictures are essential for recognizing and comprehending a wide range of medical problems. This work introduces the Direction Coupled Magnitude Histogram (DCMH) as a novel structure picture descriptor to improve diagnostic accuracy. One of DCMH's unique selling points is its ability to include the edge oriented information that are oriented in any way inside a frame, enabling the expression of delicate nuances using various gradient features. The proposed method applies cartoon texture based textural loss and DCMH based structural loss to identify and analyse structural and textural information during the denoising time. A major contribution that improves the interpretability of images by emphasizing structural aspects that is inherent to the image. The proposed DCMH_3D_GANaverage results show exceptional performance, with an SSIM of 0.972995 and PSNR of 48.74, highlighting the effectiveness of the DCMH-based method in enhancing medical picture diagnosis. The capacity of Structured Loss to improve picture interpretability and lead to a more precise diagnosis is unquestionably advantageous. The newly developed DCMH-based approach, which includes texture loss and structured components, is a promising development in healthcare image processing that will enable better patient care through enhanced diagnostic abilities.