Similarity Registration Key
Maggie Schnair
The United States National Institutes of Health (NIH) commit significant support to open-source data and software resources in order to foment reproducibility in the biomedical imaging sciences. Here, we report and evaluate a recent product of this commitment: Advanced Neuroimaging Tools (ANTs), which is approaching its 2.0 release. The ANTs open source software library consists of a suite of state-of-the-art image registration, segmentation and template building tools for quantitative morphometric analysis. In this work, we use ANTs to quantify, for the first time, the impact of similarity metrics on the affine and deformable components of a template-based normalization study. We detail the ANTs implementation of three similarity metrics: squared intensity difference, a new and faster cross-correlation, and voxel-wise mutual information. We then use two-fold cross-validation to compare their performance on openly available, manually labeled, T1-weighted MRI brain image data of 40 subjects (UCLA's LPBA40 dataset). We report evaluation results on cortical and whole brain labels for both the affine and deformable components of the registration. Results indicate that the best ANTs methods are competitive with existing brain extraction results (Jaccard=0.958) and cortical labeling approaches. Mutual information affine mapping combined with cross-correlation diffeomorphic mapping gave the best cortical labeling results (Jaccard=0.6690.022). Furthermore, our two-fold cross-validation allows us to quantify the similarity of templates derived from different subgroups. Our open code, data and evaluation scripts set performance benchmark parameters for this state-of-the-art toolkit. This is the first study to use a consistent transformation framework to provide a reproducible evaluation of the isolated effect of the similarity metric on optimal template construction and brain labeling.
Image registration is the process of transforming different sets of data into one coordinate system. Data may be multiple photographs, data from different sensors, times, depths, or viewpoints.[1] It is used in computer vision, medical imaging,[2] military automatic target recognition, and compiling and analyzing images and data from satellites. Registration is necessary in order to be able to compare or integrate the data obtained from these different measurements.
Image registration or image alignment algorithms can be classified into intensity-based and feature-based.[3] One of the images is referred to as the moving or source and the others are referred to as the target, fixed or sensed images. Image registration involves spatially transforming the source/moving image(s) to align with the target image. The reference frame in the target image is stationary, while the other datasets are transformed to match to the target.[3] Intensity-based methods compare intensity patterns in images via correlation metrics, while feature-based methods find correspondence between image features such as points, lines, and contours.[3] Intensity-based methods register entire images or sub-images. If sub-images are registered, centers of corresponding sub images are treated as corresponding feature points. Feature-based methods establish a correspondence between a number of especially distinct points in images. Knowing the correspondence between a number of points in images, a geometrical transformation is then determined to map the target image to the reference images, thereby establishing point-by-point correspondence between the reference and target images.[3] Methods combining intensity-based and feature-based information have also been developed.[4]
Image registration algorithms can also be classified according to the transformation models they use to relate the target image space to the reference image space. The first broad category of transformation models includes linear transformations, which include rotation, scaling, translation, and other affine transforms.[5] Linear transformations are global in nature, thus, they cannot model local geometric differences between images.[3]
Spatial methods operate in the image domain, matching intensity patterns or features in images. Some of the feature matching algorithms are outgrowths of traditional techniques for performing manual image registration, in which an operator chooses corresponding control points (CP) in images. When the number of control points exceeds the minimum required to define the appropriate transformation model, iterative algorithms like RANSAC can be used to robustly estimate the parameters of a particular transformation type (e.g. affine) for registration of the images.
Frequency-domain methods find the transformation parameters for registration of the images while working in the transform domain. Such methods work for simple transformations, such as translation, rotation, and scaling. Applying the phase correlation method to a pair of images produces a third image which contains a single peak. The location of this peak corresponds to the relative translation between the images. Unlike many spatial-domain algorithms, the phase correlation method is resilient to noise, occlusions, and other defects typical of medical or satellite images. Additionally, the phase correlation uses the fast Fourier transform to compute the cross-correlation between the two images, generally resulting in large performance gains. The method can be extended to determine rotation and scaling differences between two images by first converting the images to log-polar coordinates.[13][14] Due to properties of the Fourier transform, the rotation and scaling parameters can be determined in a manner invariant to translation.
Another classification can be made between single-modality and multi-modality methods. Single-modality methods tend to register images in the same modality acquired by the same scanner/sensor type, while multi-modality registration methods tended to register images acquired by different scanner/sensor types.
Multi-modality registration methods are often used in medical imaging as images of a subject are frequently obtained from different scanners. Examples include registration of brain CT/MRI images or whole body PET/CT images for tumor localization, registration of contrast-enhanced CT images against non-contrast-enhanced CT images[15] for segmentation of specific parts of the anatomy, and registration of ultrasound and CT images for prostate localization in radiotherapy.
Registration methods may be classified based on the level of automation they provide. Manual, interactive, semi-automatic, and automatic methods have been developed. Manual methods provide tools to align the images manually. Interactive methods reduce user bias by performing certain key operations automatically while still relying on the user to guide the registration. Semi-automatic methods perform more of the registration steps automatically but depend on the user to verify the correctness of a registration. Automatic methods do not allow any user interaction and perform all registration steps automatically.
Image similarities are broadly used in medical imaging. An image similarity measure quantifies the degree of similarity between intensity patterns in two images.[3] The choice of an image similarity measure depends on the modality of the images to be registered. Common examples of image similarity measures include cross-correlation, mutual information, sum of squared intensity differences, and ratio image uniformity. Mutual information and normalized mutual information are the most popular image similarity measures for registration of multimodality images. Cross-correlation, sum of squared intensity differences and ratio image uniformity are commonly used for registration of images in the same modality.
There is a level of uncertainty associated with registering images that have any spatio-temporal differences. A confident registration with a measure of uncertainty is critical for many change detection applications such as medical diagnostics.
In remote sensing applications where a digital image pixel may represent several kilometers of spatial distance (such as NASA's LANDSAT imagery), an uncertain image registration can mean that a solution could be several kilometers from ground truth. Several notable papers have attempted to quantify uncertainty in image registration in order to compare results.[16][17] However, many approaches to quantifying uncertainty or estimating deformations are computationally intensive or are only applicable to limited sets of spatial transformations.
Image registration has applications in remote sensing (cartography updating), and computer vision. Due to the vast range of applications to which image registration can be applied, it is impossible to develop a general method that is optimized for all uses.
Medical image registration (for data of the same patient taken at different points in time such as change detection or tumor monitoring) often additionally involves elastic (also known as nonrigid) registration to cope with deformation of the subject (due to breathing, anatomical changes, and so forth).[18][19][20] Nonrigid registration of medical images can also be used to register a patient's data to an anatomical atlas, such as the Talairach atlas for neuroimaging.
In cryo-TEM instability causes specimen drift and many fast acquisitions with accurate image registration is required to preserve high resolution and obtain high signal to noise images. For low SNR data, the best image registration is achieved by cross-correlating all permutations of images in an image stack.[21]
Image registration is an essential part of panoramic image creation. There are many different techniques that can be implemented in real time and run on embedded devices like cameras and camera-phones.
and I(X,Y) is the SMI given by (13) or (14). NSMI can be obtained from either SMI or SMINS. We evaluate the two new similarity measures introduced here, NSMI and NSMINS, in section 5.
dd2b598166