MRI PET images by using shearlet transform and pluse coded neural network

Abstract:

Medical image fusion plays a vital role in modern clinical diagnostics by integrating complementary information from multiple imaging modalities into a single, comprehensive representation. This paper presents a novel image fusion framework that combines Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) scans using the Non-Subsampled Shearlet Transform (NSST) and Pulse Coupled Neural Network (PCNN). MRI provides high-resolution anatomical detail of brain structures, while PET captures functional metabolic activity, making their fusion clinically significant for neurological assessment. The proposed method decomposes both input images into multi-scale, multi-directional frequency subbands using NSST, effectively capturing fine structural and textural features at each decomposition level. Low-frequency subbands are merged using an energy-attribute weighted averaging strategy to preserve global contrast and structural consistency, while high-frequency subbands are fused via PCNN, which mimics biological neural synchronization to adaptively select salient detail features. The fused subbands are then reconstructed through inverse NSST to produce the final composite image. Implemented in MATLAB, the system is evaluated using entropy, standard deviation, and Structural Similarity Index Measure (SSIM), demonstrating superior integration of anatomical and functional information for more accurate medical diagnosis and treatment planning.

Keywords: Image Fusion, Non-Subsampled Shearlet Transform (NSST), Pulse Coupled Neural Network (PCNN), Magnetic Resonance Imaging (MRI), Positron Emission Tomography (PET)