This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2019.2907769, IEEE Access VOLUME XX, 2017 1 Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000. Digital Object Identifier 10.1109/ACCESS.2017.Doi Number Distance Oriented Particle Swarm Optimizer for Brain Image Registration Chengjia Wang 1 , Member, IEEE, Keith A. Goatman 2 , James Boardman 3 , Erin Beveridge 2 , David Newby 1 , and Scott Semple 1 1 BHF Centre for Cardiovascular Science, University of Edinburgh, Edinburgh, EH16 4TJ UK 2 Canon Medical Research Europe Ltd., Edinburgh, EH6 5NP UK 3 MRC Centre for Reproductive Health, University of Edinburg, Edinburgh EH9 1UW UK Corresponding author: Chengjia Wang (e-mail: Chengjia.Wang@ed.ac.uk). This work is funded by the Medical Research Council, British Heart Foundation Centre of Research Excellence award and the Scottish Universities Physics Alliance INSPIRE award. Data were acquired with funding from Theirworld. ABSTRACT In this paper we describe improvements to the particle swarm optimizer (PSO) made by inclusion of an unscented Kalman filter to guide particle motion. We show how this increases the speed of convergence, and reduces the likelihood of premature convergence, increasing overall accuracy. We demonstrate the effectiveness of the unscented Kalman filter PSO by comparing it with the original PSO algorithm and its variants designed to improve performance. The PSOs were tested firstly on a number of common synthetic benchmarking functions, and secondly applied to a practical three-dimensional image registration problem. The proposed methods displayed better performances for 4 out of 8 benchmark functions and reduced the target registration errors by at least 2mm when registering down-sampled benchmark brain images. They also demonstrated an ability to align images featuring motion related artefacts which all other methods failed to register. These new PSO methods provide a novel, efficient mechanism to integrate prior knowledge into each iteration of the optimization process, which can enhance the accuracy and speed of convergence in the application of medical image registration. INDEX TERMS global optimization, particle swarm, unscented Kalman filter, image registration I. INTRODUCTION Optimization is a key component in many practical scientific computing problems. It is used to search for the optimum value of a pre-defined fitness function of a measure within a problem space [1]. As a typical global optimization method, particle swarm optimization (PSO) has been paid significant attention during the last few decades, as it is less prone to becoming trapped in local optima. Various improvements have been suggested to the original PSO algorithm to improve convergence and computation speed. However, neither the original PSO method nor its general- purpose modifications derived any advantage from available prior knowledge about the problem space which may act as a critical role in specific applications. The goal of many optimization problems is not just searching for an optimal value of the fitness function. One typical example of this issue is presented by a problem associated with image registration, for which the distance to the real global optima, rather than the value of the measurement function, is more important. This is because small differences of the fitness function values can actually represent large differences between image transformation parameters, which may in turn falsely indicate alignment between images. If prior knowledge about the content of the image is ignored in favour of the result of the value-oriented PSO, the optimization process may tend to converge to local optima that exhibit “better” measurement values. These local optima may be at a significant distance from the global optimum, thereby causing the image registration to “fail”. To deal with this specific application, in this paper, we introduce a novel distance-oriented PSO, guided by an unscented Kalman filter (UKF) [1]. This method can encode prior knowledge about the distribution of a fitness function within the problem space and stretch the optimizer to converge at a point near the true global optimum. Image registration algorithms are often based on the premise that the magnitude of the chosen similarity metric is related to the magnitude of the error between the current