Impact of Gaussian Noise on the Optimization of Medical Image Registration

Abstract

Gaussian noise often poses a significant challenge to medical image registration, impacting the accuracy and reliability of alignment across varying imaging modalities. The research investigates the effect of Gaussian noise on medical image registration by comparing four optimization techniques: a direct approach, an optimization using FMINCON, a multiscale approach, and a combined optimization strategy that integrates FMINCON and the multiscale approach. The comparative analysis assesses each method's robustness against Gaussian noise, evaluating registration accuracy through three key similarity metrics: Mean Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR), and Structural Similarity Index Measure (SSIM). The results reveal that while each approach demonstrates a degree of resilience to noise, the combined optimization method significantly outperforms the others, achieving the lowest MSE, highest PSNR, and superior SSIM. These findings suggest that the combined approach effectively enhances the optimization process by leveraging the strengths of both FMINCON and multiscale frameworks, thus providing a more accurate and noise-resistant solution for medical image registration. The analysis highlights the necessity of image filtering techniques to mitigate noise interference and improve the image registration process in clinical applications.