CT and MRI Medical Image Fusion Using Noise-Removal and Contrast Enhancement Scheme with Convolutional Neural Network

Abstract

Medical image fusion (MIF) has received painstaking attention due to its diverse medical applications in response to accurately diagnosing clinical images. Numerous MIF methods have been proposed to date, but the fused image suffers from poor contrast, non-uniform illumination, noise presence, and improper fusion strategies, resulting in an inadequate sparse representation of significant features. This paper proposes the morphological preprocessing method to address the non-uniform illumination and noise by the bottom-hat–top-hat strategy. Then, grey-principal component analysis (grey-PCA) is used to transform RGB images into gray images that can preserve detailed features. After that, the local shift-invariant shearlet transform (LSIST) method decomposes the images into the low-pass (LP) and high-pass (HP) sub-bands, efficiently restoring all significant characteristics in various scales and directions. The HP sub-bands are fed to two branches of the Siamese convolutional neural network (CNN) by process of feature detection, initial segmentation, and consistency verification to effectively capture smooth edges, and textures. While the LP sub-bands are fused by employing local energy fusion using the averaging and selection mode to restore the energy information. The proposed method is validated by subjective and objective quality assessments. The subjective evaluation is conducted by a user case study in which twelve field specialists verified the superiority of the proposed method based on precise details, image contrast, noise in the fused image, and no loss of information. The supremacy of the proposed method is further justified by obtaining 0.6836 to 0.8794, 0.5234 to 0.6710, and 3.8501 to 8.7937 gain for QFAB , CRR, and AG and noise reduction from 0.3397 to 0.1209 over other methods for objective parameters.

Sparse Feature Aware Noise Removal Technique for Brain Multiple Sclerosis Lesions using Magnetic Resonance Imaging

“…The Table I shows the computation outcome achieved using proposed SFANR over existing noise removal methods such as DPED [1], Morphology learning CNN (ML-CNN) [24], and noise removal using dictionary learning (NRDL) [1]. The SFANR achieves much lesser running time than other existing methodologies; thus, are very efficient.…”

“…The dictionary is constructed leveraging Maximum posterior combined using noisy MRI. Similarly, [24] used morphological features to address the impact of noise due to non-uniform illumination. Further, employed principal component analysis retains detailed features such as textures, smooth edges, etc.…”

“…The CNN is widely used for removing noise from MRI and CT [19], [20]. The CNN-based noise removal techniques [21], [22], provide an effective way of eliminating a different kind of noise [23] from MRI and improving its reconstruction quality [24], [25], and enhancing segmentation [26], [27] and classification outcomes as well [28], [29]. The CNN employs a hierarchical learning mechanism where a feed-forward network is used for the extraction of different features and a hidden layer is used to optimize the feature learning weights.…”

“…However, weights of a convolutional network [22] are updated individually for the entire MRI information; thus, the final reconstructed image exhibit higher noise, poor contrast, and limited sparse feature representation. In [23] addressed the above issues by employing morphology-based [24] feature construction using CNN. The model is efficient in eliminating noise with very good reconstruction quality; however, there exist optimization constraints in obtaining good quality sparse features.…”

Medical Image Fusion Using CNN with Automated Pooling

“…Morphological pre-processing method for noise reduction (15) presents the bottom-hat-top-hat technique as a morphological pre-processing method to deal with noise and non-uniform lighting. Then, RGB images are converted into grayscale images that may retain fine details using grey-PCA.…”