Medical Image Fusion Based On Hybrid Intelligence and Local Energy In The Nonsubsampled Shearlet Domain

Multimodality image fusion has been shown to be effective in a variety of therapeutic settings where clinicians require image support. Several medical computer-assisted diagnosis systems have recently been updated to include fusion algorithms for real-time decision making in image guided interventions and radiation treatment planning. Medical images are prone to uncertainty and ambiguity. To this end fuzzy systems can typically deal with these uncertainties. We introduce a hybrid intelligence and multi-scale decomposition (MSD) based multimodality medical image fusion algorithm in this study. The Non-subsampled shearlet transform (NSST) first divides the input images into high and low frequency subbands. The high frequency coefficients are then fused at each level of decomposition using the neuro-fuzzy inference system, while the low frequency coefficients are mixed using the local energy. Experiments on pre-registered CT and MR medical imaging datasets were carried out to evaluate the performance of the proposed approach. The proposed technique outperforms standard MSD tools by preserving more edge information, and contrast according to experimental data. Our proposed methodology transfers crucial visual information and contrast according to some quantitative parameters, which is superior to state-of-the-art methods.