This paper presents a robust and inverse consistent registration method for neuroimaging data, which is crucial for quantifying changes in brain structures over time. The method is designed to handle various challenges such as differential distortions, true anatomical changes, and motion artifacts. Inspired by Nestares and Heeger (2000), the approach uses robust statistics to discount regions with true differences and recover the correct alignment. Key features include guaranteeing inverse consistency, handling different intensity scales, and automatically estimating a sensitivity parameter to detect outliers. The method outperforms state-of-the-art tools like FLIRT and SPM in terms of symmetry, robustness, and accuracy. The paper also discusses the theoretical background, transformation models, and intensity scaling, and provides a detailed algorithm description. Experimental results demonstrate the method's superior performance in both synthetic and real data, particularly in longitudinal studies where motion correction and averaging of intra-session scans are essential.This paper presents a robust and inverse consistent registration method for neuroimaging data, which is crucial for quantifying changes in brain structures over time. The method is designed to handle various challenges such as differential distortions, true anatomical changes, and motion artifacts. Inspired by Nestares and Heeger (2000), the approach uses robust statistics to discount regions with true differences and recover the correct alignment. Key features include guaranteeing inverse consistency, handling different intensity scales, and automatically estimating a sensitivity parameter to detect outliers. The method outperforms state-of-the-art tools like FLIRT and SPM in terms of symmetry, robustness, and accuracy. The paper also discusses the theoretical background, transformation models, and intensity scaling, and provides a detailed algorithm description. Experimental results demonstrate the method's superior performance in both synthetic and real data, particularly in longitudinal studies where motion correction and averaging of intra-session scans are essential.