A full migration BBO algorithm with enhanced population quality bounds for multimodal biomedical image registration

Highlights

• A novel BBO algorithm: the Biogeography-based Optimization algorithm with Elite Learning (BBO-EL).

• A hybrid full migration operator to perform the migration operation and increase the search space.

• The lower bound of the whole population’s quality can be maintained in higher level.

• An elite learning operator based on social comparison theory to improve the upper bound.

Abstract

Medical images acquired from different modalities give rise to many practical problems in image registration. Intensity-based registration techniques have been increasingly used in multimodal image registration; these techniques integrate different images that have shared content into a single representation, by transformation. The estimation of the optimal transformation requires the optimization of a similarity metric between the images. Recently, many optimization methods have been proposed that focus on the development of the optimization component. However, there is still room for large amounts of improvement, from both an efficiency point of view and a quality perspective. In this paper we present a new Biogeography-based Optimization (BBO) algorithm, the Biogeography-based Optimization algorithm with Elite Learning (BBO-EL), for multimodal medical image registration. First, we propose a hybrid full migration operator in which each individual has the chance to perform the migration operation and the whole population has the chance to expand the search space. In this way, the search ability of the BBO algorithm is enhanced and matches well the characteristics of multimodal medical image registration. In addition, considering that the quality of some individuals could be deteriorated as caused by the migration operation, we propose an undo operator on the deteriorated individuals. Thus, the lower bound of the whole population’s quality can be maintained at a higher level. Furthermore, in the original BBO algorithm, a number of good individuals might be not involved in the migration operation, and we present an elite learning operator that is based on social comparison theory to improve the upper bound of the whole population’s quality. Therefore, after improving both the lower bound and the upper bound of the whole population’s quality, the accuracy and the convergence speed of the multimodal medical registration can be greatly enhanced. The BBO-EL has been tested in many experiments on benchmark datasets include six kind of different modality images, from up to eighteen different patients, which can make up 54 multimodal registration scenarios. The BBO-EL obtained 30 best performance scenarios while the state-of-the-art algorithm obtained 21 scenarios. The results demonstrated that BBO-EL outperforms the state-of-the-art algorithm in most cases for practical problems.

Introduction

Medical imaging is the technique and process of creating visual representations of the interior of a body for diagnosis, disease monitoring and medical intervention, as well as visual representation of the function of some organs or tissues [1], [2], [3]. In the development of image acquisition techniques, the images are acquired from different devices for different representations [4], [5]. For example, computer-assisted tomography (CT) is very helpful for imaging bodily structures and dense tissues, while magnetic resonance imaging (MRI) and ultrasound (US) are powerful tools for the visualization of soft tissue. In addition, functional imaging is becoming more and more important in clinical and medical research. For example, positron emission tomography (PET) and single photon computed tomography (SPECT) imaging provide information about blood flow and metabolic processes. In most cases, many of the applications address images with different modalities that were acquired from different acquisition devices at different times, possibly from different perspectives for image comparison, integration, or the fusion of visualizations. These multimodal images are used in a complementary approach to acquire additional insights into a pathological phenomenon. For these multimodal images to be used, they must be properly combined or fused. Therefore, registration is a key technology in medical imaging, as it enables us to align and integrate different images that have a shared content, which are obtained under different conditions, into a single representation [6].

Medical image registration is becoming increasingly important in diagnosis, treatment planning, functional studies, computer-guided therapies, and medical research. Registration involves the transformation of different images that have a shared content into a common coordinate system. Specifically, the objective of the registration is to estimate the optimal geometric transformation that maps the content of one image to the corresponding area of the other. For image processing [7], [8], [9], [10], [11], one key issue of registration is to find the optimal transformation that is settled by optimization methods guided by a similarity metric [12], [13] that can measure the degree of resemblance between two images.

There are many methods for medical image registration. These methods can be roughly divided into two categories: feature-based and intensity-based. Feature-based methods [6] perform alignments using only parts of the image that are salient and distinctive parts, such as lines, corners and contours. These methods can greatly reduce the complexity of the registration problem, while they can significantly depend on the reliability of the feature extraction algorithm. However, these methods are limited to the images that have salient and distinctive features. On the other hand, much work has recently focused on intensity-based methods, in which all of the image data is utilized. A large amount of data can be processed at the expense of increasing the computational requirements, which does not require the extra preprocessing, such as segmentation or feature extraction, in the intensity-based methods. There are usually three important aspects to be considered [14], [15]. First, the search domain is the group of potential transformations that are used to align the images. Second, the similarity metric is a real valued function that quantifies the degree of resemblance between two images. Finally, the search method explores the domain of possible transformations guided by the similarity metric. There are local and global search methods to optimize the similarity metric.

As is known, the key to obtaining good registration results is to find a suitable search method that can optimize the similarity metric. Therefore, the motivation of our work is to design an optimization algorithm to maximize the similarity metric for aligning medical images that are obtained from different modalities. Many algorithms have been proposed in the medical image registration area. For example, the original Powell’s method and its recent variants [16], discrete optimization of Markov random fields (MRFs) [17] and different gradient-based approaches [18] are distinguished from traditional optimization algorithms. However, many adverse factors, such as noise, image discretization and misalignment affect the process of traditional optimization algorithms, which can easily become stuck at local optima. To overcome some of these drawbacks, many metaheuristics-based methods [19] have been proposed and demonstrated. Metaheuristics-based methods have successful applications in many fields [20], [21], [22], [23] and are often used to appropriately settle complex medical image registration problems, due to their advantages in obtaining a robust global optimum. Many evolutionary methods have been proposed in recent decades with the purpose of improving the results of the different medical image registration problems [24], [25]. Among them, an approach based on the Coral Reef Optimization (CRO) algorithm, which is called CRO-SL, was proposed with outstanding results [26]. As far as we know, CRO-SL is the most notable and represents the state-of-the-art. The biogeography-based optimization algorithm (BBO) [27] is a search algorithm that is based on the law of biological migration [28]. This algorithm has good convergence and stability and is widely used in various fields. In this work, we design a new BBO based optimization that is specifically tailored to the medical image registration problem.