Effective Image Fusion of PET-MRI Brain Images Using Wavelet Transforms

2 Literature Review

Brundashree .R et al (2015) combined PET images and MRI images from the brain are utilizing wavelet transforms for tumor detection. After applying wavelet decomposition technique and fusion in gray level acquire better result. This is done by modifying basic data in Gray Matter (GM) region and fixing the data in the zone of white Matter (WM). Authors utilize three different datasets: normal axial data file, normal coronal dataset and the dataset from the brain disease Alzheimer’s for checking and evaluation. The exhibition of the combination of spectral discrepancy (SD) strategy and the value of average gradient (AG) strategy provides a proof to be enhanced result with numerically and visually. The image fusion helps to collect valued data from various pictures into form a single image. The mixture of PET and MRI brain disease images for the current concept creates the better normal tendency by which the spatial aim of picture is elevated. The spectral discrepancy for the dataset 1 and 3 produces same result as the current strategy, dataset 2 is changes to bring downwards the qualities. Henceforth the spatial aim of the picture are improved when using with the proposed technique .

Breast tissues material parameters and direction of the gravitational forces are estimated by using biomechanical model. Remaining deformation details are finding out by using non-rigid intensity based image registration.

Introduced A-NSIFT algorithm enhances the interest key point extraction procedure. Compare to conventional NSIFT, A –NSIFT is 200 time faster in interest point extraction. Further key point matching a probability density estimation Gaussian mixture model is used for alignment. From the experimental result it is concluded that the given module presents the faster 3D alignment even though there is poor initialization.

But the most challenging thing is to integrate or fuse two different medical images. The referred paper propose an efficient algorithm to integrated both MR and ultrasound images. Robust PaTch-based correlations Ratio (RaPTOR) used to register the 3D volumetric data of both ultrasound and MR data. The referred system proposed a decent approach towards the image registration module which speeds up the performance.

A similarity based image registration module in which correlation among the pixel is computed by Sum-of-Squared-Difference (SSD) is presented. Along with SSD the aid of the low rank matrix theory is applied. Application of the theory will compensate the intensity distortion which leads to the rank regularised SSD (RRSSD). The proposed architectures effectively perform intensity distortion correction and image registration at a time. Experimental result concluded that designed algorithm effectively meets the clinically acceptable image registration output.