Optimised MRI intensity standardisation based on multi-dimensional sub-regional point cloud registration

As one of the most widely used medical imaging methods, magnetic resonance imaging (MRI) and its corresponding automatic image analysis are research hotspots in computer-aided medical diagnosis. However, due to the disparities in imaging parameters and scanner characteristics, the variation in intensity distribution between scanners results in performance reduction in automatic image analysis and diagnosis. This paper aims at obtaining sub-region intensity distribution and forming a robust non-rigid intensity transforming function in order to standardise the intensities of MR images. The proposed method includes multi-modality image registration, sub-region standard intensity estimation, weighted -dimensional point cloud generation and global intensity transformation. Between the target images and reference images, this novel method could not only avoid the intensity distortion due to the inconsistency of inter-tissue brightness relationship, but also reduce the dependence on the accuracy of multi-modality MR image registration. Experiments show performance enhancement in peak signal-to-noise ratio, average disparity, histogram correlation, mean square error and structural similarity over the existing methods. Therefore, the intensities of MR images from different scanners could be standardised by the proposed method, so that multi-centre/multi-machine correlation could be promoted.