A Novel Medical Image Fusion Approach Using Local Features And Deep Learning In Wavelet Domain

Abstract

In clinical evaluation, medical image analysis is essential. Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) are widely utilized in the medical sector for disease diagnosis. However, extracting the necessary information to detect suspicious tissue features from a single image is quite challenging. Multimodal image fusion has garnered increased interest because it allows physicians to view multiple modalities in one image. Physicians can accurately diagnose diseases and effectively plan treatment by analyzing the fused image. Numerous techniques have been devoted to image fusion. However, the fused images have issues with brightness, contrast, and noise, making it difficult to precisely analyze the images. To tackle these issues, this paper introduces a new Medical Image Fusion (MIF) method based on Wavelet Transform (WT) and Deep Convolutional Neural Network (DCNN). The key objective of the proposed method is to maintain diagnostic data in the fused image. The introduced method applies WT on source images to divide approximate and detail subbands. Approximation subbands are fused using local features. Detail subbands are denoised using an improved thresholding method and then fused utilizing DCNN. After fusion, inverse WT is employed to obtain the final fused image. MRI and CT scans from the The effectiveness of the suggested approach is verified using Brain Atlas.  Experimental outcomes highlight the superior performance of the proposed model when compared to recent medical image fusion approaches.