Fourier registration of three-dimensional brain MR images: Exploiting the axis of rotation

We have developed a novel algorithm to register three-dimensional MR images that have undergone rigid body motion. The most interesting feature of the algorithm is that it reduces a general three-dimensional rotation to a simple planar rotation by finding the axis of rotation. The algorithm, which is a nontrivial three-dimensional extension of existing Fourier registration algorithms, has been tested on 30 artificially misaligned MR images of a phantom, four artificially misaligned MR images of a brain, and one case of actual patient motion. The algorithm successfully registered every image. The registration error for a voxel 10 cm from the origin for the artificially misaligned phantom images was 2.8 mm at most and had a mean of 1.2 mm and standard deviation of .7 mm. The registration parameters for the images contaminated by actual patient motion were similar to that from an established image registration algorithm. The results indicate that the algorithm is accurate, reliable, and fast. The rigid body model requires the brain to be segmented from MR images of the head before registration.     

Automatic method to assess local CT-MR imaging registration accuracy on images of the head

BACKGROUND AND PURPOSE: Precise registration of CT and MR images is crucial in many clinical cases for proper diagnosis, decision making or navigation in surgical interventions. Various algorithms can be used to register CT and MR datasets, but prior to clinical use the result must be validated. To evaluate the registration result by visual inspection is tiring and time-consuming. We propose a new automatic registration assessment method, which provides the user a color-coded fused representation of the CT and MR images, and indicates the location and extent of poor registration accuracy. METHODS: The method for local assessment of CT-MR registration is based on segmentation of bone structures in the CT and MR images, followed by a voxel correspondence analysis. The result is represented as a color-coded overlay. The algorithm was tested on simulated and real datasets with different levels of noise and intensity non-uniformity. RESULTS: Based on tests on simulated MR imaging data, it was found that the algorithm was robust for noise levels up to 7% and intensity non-uniformities up to 20% of the full intensity scale. Due to the inability to distinguish clearly between bone and cerebro-spinal fluids in the MR image (T1-weighted), the algorithm was found to be optimistic in the sense that a number of voxels are classified as well-registered although they should not. However, nearly all voxels classified as misregistered are correctly classified. CONCLUSION: The proposed algorithm offers a new way to automatically assess the CT-MR image registration accuracy locally in all the areas of the volume that contain bone and to represent the result with a user-friendly, intuitive color-coded overlay on the fused dataset.     

First-pass myocardial perfusion image registration by maximization of normalized mutual information

PURPOSE: To evaluate a left ventricular image registration algorithm for first-pass MR myocardial perfusion. MATERIALS AND METHODS: A normalized mutual information based motion correction algorithm was proposed and tested on 27 adenosine stressed myocardial perfusion cases consisting of pretreatment and posttreatment of 15 patients undergone autologous bone marrow mononuclear cell transplant therapy. An image mask approximately covering the left and right ventricles was manually defined to include a region of interest for registration. A two-dimensional multiresolution registration approach was used to register consecutively acquired multislice images with in-plane translations. The method was validated by manual registration and singular value deconvolution based perfusion analysis. RESULTS: The proposed image registration algorithm was found to be robust in minimizing the in-plane motion of the left ventricle in first-pass myocardial perfusion. The image mask including the left and right ventricle was found to be more robust than including the left ventricle alone. A smooth estimate of normalized mutual information coefficients were achieved for images with large contrast changes. CONCLUSION: The proposed semiautomatic multiresolution registration algorithm was able to register first-pass MR myocardial perfusion images and may be useful in quantitative perfusion analysis.     

Deformation invariant attribute vector for deformable registration of longitudinal brain MR images

This paper presents a novel approach to define deformation invariant attribute vector (DIAV) for each voxel in 3D brain image for the purpose of anatomic correspondence detection. The DIAV method is validated by using synthesized deformation in 3D brain MRI images. Both theoretic analysis and experimental studies demonstrate that the proposed DIAV is invariant to general nonlinear deformation. Moreover, our experimental results show that the DIAV is able to capture rich anatomic information around the voxels and exhibit strong discriminative ability. The DIAV has been integrated into a deformable registration algorithm for longitudinal brain MR images, and the results on both simulated and real brain images are provided to demonstrate the good performance of the proposed registration algorithm based on matching of DIAVs.     

Behaviors of cost functions in image registration between <Superscript>201</Superscript>Tl brain tumor single-photon emission computed tomography and magnetic resonance images

Objective  We studied the behaviors of cost functions in the registration of thallium-201 (²⁰¹Tl) brain tumor single-photon emission computed tomography (SPECT) and magnetic resonance (MR) images, as the similarity index of image positioning. Methods  A marker for image registration [technetium-99m (^99mTc) point source] was attached at three sites on the heads of 13 patients with brain tumor, from whom 42 sets of ^99mTc-²⁰¹Tl SPECT (the dual-isotope acquisition) and MR images were obtained. The ²⁰¹Tl SPECT and MR images were manually registered according to the markers. From the positions where the two images were registered, the position of the ²⁰¹Tl SPECT was moved to examine the behaviors of the three cost functions, i.e., ratio image uniformity (RIU), mutual information (MI), and normalized MI (NMI). Results  The cost functions MI and NMI reached the maximum at positions adjacent to those where the SPECT and MR images were manually registered. As for the accuracy of image registration in terms of the cost functions MI and NMI, on average, the images were accurately registered within 3° of rotation around the X-, Y-, and Z-axes, and within 1.5 mm (within 2 pixels), 3 mm (within 3 pixels), and 4 mm (within 1 slice) of translation to the X-, Y-, and Z-axes, respectively. In terms of rotation around the Z-axis, the cost function RIU reached the minimum at positions where the manual registration of the two images was substantially inadequate. Conclusions  The MI and NMI were suitable cost functions in the registration of ²⁰¹Tl SPECT and MR images. The behavior of the RIU, in contrast, was unstable, being unsuitable as an index of image registration.     

Prospective registration of human head magnetic resonance images for reproducible slice positioning using localizer images

PURPOSE: To facilitate assessing brain tumor growth and progression of stroke lesions by reproducible slice positioning in human head magnetic resonance (MR) images, a method for prospective registration is proposed that adjusts the image slice position without moving the patient and with no additional scans. MATERIALS AND METHODS: The gradient reference frame of follow-up examinations was adjusted to achieve the same image slice positioning relative to the patient as in the previous examination. The three-dimensional geometrical transformation parameters for the gradients were determined using two-dimensional image registration of three orthogonal localizer images. The method was developed and evaluated using a phantom with arbitrarily adjustable position. Feasibility for in vivo applications was demonstrated with brain MR imaging (MRI) of healthy volunteers. RESULTS: Standard retrospective registration was used for assessing the quality of the method. The accuracy of the realignment was 0.0 mm +/- 1.2 mm and -0.2 degrees +/- 0.9 degrees (mean +/- SD) in phantom experiments. In 10 examinations of volunteers, misalignments up to 49.2 mm and 21 degrees were corrected. The accuracy of the realignment after prospective registration was 0.1 mm +/- 1.5 mm and 0.2 degrees +/- 1.5 degrees. CONCLUSION: Image-based prospective registration using localizer images of the pre- and postexaminations is a robust method for reproducible slice positioning.     

Image content extraction: application to MR images of the brain.

A system for automatically extracting image content features was developed that combines registration to a labeled atlas with natural language processing of free-text radiology reports. The system was then tested with T1-weighted, spoiled gradient-echo magnetic resonance (MR) imaging studies of the brain performed in nine patients. The locations of 599 structures were visually assessed by an experienced radiologist and compared with the locations indicated by automated output. The in-plane accuracy of the contours was subjectively evaluated as either good, moderate, or poor. The criterion for classifying a structure as correctly located was that 90% or more of all the images containing the structure had to be correctly identified. For 98% of the structures, the images identified by the automated algorithm agreed with those identified by the radiologist, and in 83% of cases, image contours showed a good in-plane overlap. The results of this validation study demonstrate that this combination of registration and natural language processing is accurate in identifying relevant images from brain MR imaging studies. However, the range of applicability of this technique has yet to be determined by applying the technique to a large number of studies.     

Registration of three-dimensional MR and PET images of the human brain without markers

Axial and sagittal magnetic resonance (MR) sections and contiguous sections of axial positron emission tomographic (PET) images obtained with fludeoxyglucose F-18 were used to evaluate a new method of registering three-dimensional images of the brain. The users specified the interhemispheric fissure plane in three dimensions for both the MR and PET data sets by specifying its endpoints within several axial sections. A transformation matrix aligning the interhemispheric fissure plane in MR and PET space was calculated and used to create one resectioned PET image on the resectioned PET image, and the user specified the remaining translations and rotation by moving the overlaid outline of the MR image. MR and PET data sets in four subjects were registered. The three-dimensional error on average was less than 3.8 mm and never exceeded 7.5 mm. Less than 1 hour per patient was required for registration. The method is accurate unless the interhemispheric fissure deviates significantly from a planar configuration. It does not need thin or contiguous MR sections and provides an estimate of the total registration error for every case.     

Magnetic resonance image registration in multiple sclerosis: comparison with repositioning error and observer-based variability

PURPOSE: To study the use of image registration in the analysis of multiple sclerosis (MS) lesion volume and compare this with repositioning error and observer-based variability. MATERIALS AND METHODS: The normalized mutual information (NMI) algorithm is evaluated in an accuracy study using a phantom, followed by a validation study on magnetic resonance (MR) data of MS patients. Further, using scan-rescan MR data, the effect of registration on MS lesion volume compared with repositioning error and observer-based variability is assessed. RESULTS: The registration accuracy was near perfect in the phantom study, while the in vivo validation study demonstrated an accuracy on the order of 0.2-0.3 mm. In the scan-rescan study, quantification accounted for 15.6% of the relative variance, repositioning for 44.4%, and registration for 40.0%. CONCLUSION: NMI resulted in robust and accurate alignment of MR brain images of MS patients. Its use in the detection of changes in MS using large serial MR imaging (MRI) data warrants future evaluation.     

PET-MR image fusion in soft tissue sarcoma: accuracy,reliability and practicality of interactive point-based and automated mutual information techniques

The fusion of functional positron emission tomography (PET) data with anatomical magnetic resonance (MR) or computed tomography images, using a variety of interactive and automated techniques, is becoming commonplace, with the technique of choice dependent on the specific application. The case of PET-MR image fusion in soft tissue is complicated by a lack of conspicuous anatomical features and deviation from the rigid-body model. Here we compare a point-based external marker technique with an automated mutual information algorithm and discuss the practicality, reliability and accuracy of each when applied to the study of soft tissue sarcoma. Ten subjects with suspected sarcoma in the knee, thigh, groin, flank or back underwent MR and PET scanning after the attachment of nine external fiducial markers. In the assessment of the point-based technique, three error measures were considered: fiducial localisation error (FLE), fiducial registration error (FRE) and target registration error (TRE). FLE, which represents the accuracy with which the fiducial points can be located, is related to the FRE minimised by the registration algorithm. The registration accuracy is best characterised by the TRE, which is the distance between corresponding points in each image space after registration. In the absence of salient features within the target volume, the TRE can be measured at fiducials excluded from the registration process. To assess the mutual information technique, PET data, acquired after physically removing the markers, were reconstructed in a variety of ways and registered with MR. Having applied the transform suggested by the algorithm to the PET scan acquired before the markers were removed, the residual distance between PET and MR marker-pairs could be measured. The manual point-based technique yielded the best results (RMS TRE =8.3 mm, max =22.4 mm, min =1.7 mm), performing better than the automated algorithm (RMS TRE =20.0 mm, max =30.5 mm, min =7.7 mm) when registering filtered back-projection PET images to MR. Image reconstruction with an iterative algorithm or registration of a composite emission-transmission image did not improve the overall accuracy of the registration process. We have demonstrated that, in this application, point-based PET-MR registration using external markers is practical, reliable and accurate to within approximately 5 mm towards the fiducial centroid. The automated algorithm did not perform as reliably or as accurately.     

Automatic registration of MR and SPECT images for treatment planning in prostate cancer

RATIONALE AND OBJECTIVES: To aid in surgical and radiation therapy planning for prostate adenocarcinoma, a general-purpose automatic registration method that is based on mutual information was used to align magnetic resonance (MR) images and single photon emission computed tomographic (SPECT) images of the pelvis and prostate. MATERIALS AND METHODS: The authors assessed the effects of various factors on alignment between pairs of MR and SPECT images, including the use of particular pulse sequences in MR imaging, image voxel intensity scaling, the use of different regions on the MR-SPECT histogram, spatial masking of nonoverlapping visual data between images, and multiresolution optimization. A mutual information algorithm was used as the cost function for automatic registration. Automatic registration was deemed acceptable when it resulted in a transformation with less than 2 voxel units (6 mm) difference in translation and less than 2 degree difference in rotation from that obtained with manual registration performed independently by nuclear medicine radiologists. RESULTS: Paired sets of MR and SPECT image volumes from four of five patients were successfully registered. For successful registration, MR images must be optimal and registration must be performed at full spatial resolution and at the full intensity range. Masking, cropping, and the normalization of mutual information, used to register partially overlapping MR-SPECT volumes, were not successful. Multiresolution optimization had little effect on the accuracy and speed of the registration. CONCLUSION: Automatic registration between MR and SPECT images of the pelvis can be achieved when data acquisition and image processing are performed properly. It should prove useful for prostate cancer diagnosis, staging, and treatment planning.     

Automatic removal of extrameningeal tissues from MR images of human brain

An automatic technique for removal of extrameningeal tissues from MR images of human brain is described. The algorithm is based on the segmentation of images acquired with a fast dual-echo pulse sequence, which incorporates both fluid attenuation and magnetization transfer contrast for superior brain/extrameningeal tissue contrast compared to conventional MR sequences. Evaluation of this technique using MR images of six volunteers indicates rapid and consistent removal of extrameningeal tissues.     

A validation framework for brain tumor segmentation

RATIONALE AND OBJECTIVES: We introduce a validation framework for the segmentation of brain tumors from magnetic resonance (MR) images. A novel unsupervised semiautomatic brain tumor segmentation algorithm is also presented. MATERIALS AND METHODS: The proposed framework consists of 1) T1-weighted MR images of patients with brain tumors, 2) segmentation of brain tumors performed by four independent experts, 3) segmentation of brain tumors generated by a semiautomatic algorithm, and 4) a software tool that estimates the performance of segmentation algorithms. RESULTS: We demonstrate the validation of the novel segmentation algorithm within the proposed framework. We show its performance and compare it with existent segmentation. The image datasets and software are available at http://www.brain-tumor-repository.org/. CONCLUSIONS: We present an Internet resource that provides access to MR brain tumor image data and segmentation that can be openly used by the research community. Its purpose is to encourage the development and evaluation of segmentation methods by providing raw test and image data, human expert segmentation results, and methods for comparing segmentation results.     

Validation of fully automatic brain SPET to MR co-registration

Fully automatic co-registration of functional to anatomical brain images using information intrinsic to the scans has been validated in a clinical setting for positron emission tomography (PET), but not for single-photon emission tomography (SPET). In this paper we evaluate technetium-99m hexamethylpropylene amine oxime to magnetic resonance (MR) co-registration for five fully automatic methods. We attached six small fiducial markers, visible in both SPET and MR, to the skin of 13 subjects. No increase in the radius of SPET acquisition was necessary. Distortion of the fiducial marker distribution observed in the SPET and MR studies was characterised by a measure independent of registration and three subjects were excluded on the basis of excessive distortion. The location of each fiducial marker was determined in each modality to sub-pixel precision and the inter-modality distance was averaged over all markers to give a fiducial registration error (FRE). The component of FRE excluding the variability inherent in the validation method was estimated by computing the error transformation between the arrays of MR marker locations and registered SPET marker locations. When applied to the fiducial marker locations this yielded the surface registration error (SRE), and when applied to a representative set of locations within the brain it yielded the intrinsic registration error (IRE). For the best method, mean IRE was 1.2 mm, SRE 1.5 mm and FRE 2.4 mm (with corresponding maxima of 3.3, 4.3 and 5.0 mm). All methods yielded a mean IRE <3 mm. The accuracy of the most accurate fully automatic SPET to MR co-registration was comparable with that published for PET to MR. With high standards of calibration and instrumentation, intra-subject cerebral SPET to MR registration accuracy of <2 mm is attainable.     

Validation of fully automatic brain SPET to MR co-registration

Fully automatic co-registration of functional to anatomical brain images using information intrinsic to the scans has been validated in a clinical setting for positron emission tomography (PET), but not for single-photon emission tomography (SPET). In this paper we evaluate technetium-99m hexamethylpropylene amine oxime to magnetic resonance (MR) co-registration for five fully automatic methods. We attached six small fiducial markers, visible in both SPET and MR, to the skin of 13 subjects. No increase in the radius of SPET acquisition was necessary. Distortion of the fiducial marker distribution observed in the SPET and MR studies was characterised by a measure independent of registration and three subjects were excluded on the basis of excessive distortion. The location of each fiducial marker was determined in each modality to sub-pixel precision and the inter-modality distance was averaged over all markers to give a fiducial registration error (FRE). The component of FRE excluding the variability inherent in the validation method was estimated by computing the error transformation between the arrays of MR marker locations and registered SPET marker locations. When applied to the fiducial marker locations this yielded the surface registration error (SRE), and when applied to a representative set of locations within the brain it yielded the intrinsic registration error (IRE). For the best method, mean IRE was 1.2 mm, SRE 1.5 mm and FRE 2.4 mm (with corresponding maxima of 3.3, 4.3 and 5.0 mm). All methods yielded a mean IRE <3 mm. The accuracy of the most accurate fully automatic SPET to MR co-registration was comparable with that published for PET to MR. With high standards of calibration and instrumentation, intra-subject cerebral SPET to MR registration accuracy of <2 mm is attainable. Received 29 May and in revised form 6 October 1999     

Comparison and evaluation of rigid, affine, and nonrigid registration of breast MR images.

PURPOSE: A new nonrigid registration method, designed to reduce the effect of movement artifact in subtraction images from breast MR, is compared with existing rigid and affine registration methods. METHOD: Nonrigid registration was compared with rigid and affine registration methods and unregistered images using 54 gadolinium-enhanced 3D breast MR data sets. Twenty-seven data sets had been previously reported normal, and 27 contained a histologically proven carcinoma. The comparison was based on visual assessment and ranking by two radiologists. RESULTS: When analyzed by two radiologists independently, all three registration methods gave better-quality subtraction images than unregistered images (p < 0.01), but nonrigid registration gave significantly better results than the rigid and affine registration methods (p < 0.01). There was no significant difference between rigid and affine registration methods. CONCLUSION: Nonrigid registration significantly reduces the effects of movement artifact in subtracted contrast-enhanced breast MRI. This may enable better visualization of small tumors and those within a glandular breast.     

MRI-based attenuation correction for PET/MRI: a novel approach combining pattern recognition and atlas registration

For quantitative PET information, correction of tissue photon attenuation is mandatory. Generally in conventional PET, the attenuation map is obtained from a transmission scan, which uses a rotating radionuclide source, or from the CT scan in a combined PET/CT scanner. In the case of PET/MRI scanners currently under development, insufficient space for the rotating source exists; the attenuation map can be calculated from the MR image instead. This task is challenging because MR intensities correlate with proton densities and tissue-relaxation properties, rather than with attenuation-related mass density. METHODS: We used a combination of local pattern recognition and atlas registration, which captures global variation of anatomy, to predict pseudo-CT images from a given MR image. These pseudo-CT images were then used for attenuation correction, as the process would be performed in a PET/CT scanner. RESULTS: For human brain scans, we show on a database of 17 MR/CT image pairs that our method reliably enables estimation of a pseudo-CT image from the MR image alone. On additional datasets of MRI/PET/CT triplets of human brain scans, we compare MRI-based attenuation correction with CT-based correction. Our approach enables PET quantification with a mean error of 3.2% for predefined regions of interest, which we found to be clinically not significant. However, our method is not specific to brain imaging, and we show promising initial results on 1 whole-body animal dataset. CONCLUSION: This method allows reliable MRI-based attenuation correction for human brain scans. Further work is necessary to validate the method for whole-body imaging.     

A comparative study on manual and automatic slice-to-volume registration of CT images

In order to assess the clinical relevance of a slice-to-volume registration algorithm, this technique was compared to manual registration. Reformatted images obtained from a diagnostic CT examination of the lower abdomen were reviewed and manually registered by 41 individuals. The results were refined by the algorithm. Furthermore, a fully automatic registration of the single slices to the whole CT examination, without manual initialization, was also performed. The manual registration error for rotation and translation was found to be 2.7±2.8 ° and 4.0±2.5 mm. The automated registration algorithm significantly reduced the registration error to 1.6±2.6 ° and 1.3±1.6 mm (p = 0.01). In 3 of 41 (7.3%) registration cases, the automated registration algorithm failed completely. On average, the time required for manual registration was 213±197 s; automatic registration took 82±15 s. Registration was also performed without any human interaction. The resulting registration error of the algorithm without manual pre-registration was found to be 2.9±2.9 ° and 1.1±0.2 mm. Here, a registration took 91±6 s, on average. Overall, the automated registration algorithm improved the accuracy of manual registration by 59% in rotation and 325% in translation. The absolute values are well within a clinically relevant range.     

Magnetization transfer or spin-lock? An investigation of off-resonance saturation pulse imaging with varying frequency offsets.

PURPOSETo characterize near-resonance saturation pulse MR imaging on a 1.5-T scanner in order to gain insight into underlying mechanisms that alter tissue contrast and to optimize the technique for neuroimaging.METHODSOff-resonance saturation pulses were applied to T1-weighted, spin-density-weighted, and T2-weighted sequences at frequency offsets ranging from 50 Hz to 20,000 Hz down field from water resonance. Suppression ratios were determined at each offset for phantom materials (MnCl2 solution, gadopentetate dimeglumine, corn oil, water, and agar), normal brain structures, and a variety of brain lesions.RESULTSSignal suppression of MnCl2 on T1-weighted images occurred at offsets of less than 2000 Hz even though no macromolecules were present in the solution. Only those phantom materials and tissues with short or intermediate T1 relaxation times and relatively large T1/T2 ratios were sensitive to changing frequency offsets. Suppression of brain increased from approximately 20% at 2000 Hz offset to approximately 45% when the offset was reduced to 300 Hz. In human subjects, the net effect of reducing the frequency offset was to increase T2 contrast on T1-weighted, spin-density-weighted, and T2-weighted images. Distilled water and contrast material did not suppress except at very low offsets ( < 300 Hz). A frequency offset of 300 Hz was optimal for maximizing conspicuity between most contrast-enhancing lesions and adjacent brain while preserving anatomic detail.CONCLUSIONSuppression of MnCl2 indicates that magnetization transfer is not the sole mechanism of contrast in near-resonance saturation MR imaging. Spin-lock excitation can reasonably explain the behavior of the phantom solutions and the increase in T2 contrast of tissues achieved as the frequency offset is decreased from 2000 Hz to 300 Hz. Below 300 Hz, saturation is presumably caused by spin-tip effects. With our pulse design, an offset of 300 Hz is optimal for many routine clinical imaging examinations.     

Transmission imaging for registration of ictal and interictal single-photon emission tomography, magnetic resonance imaging and electroencephalography

A method developed for registration of ictal and interictal single-photon emission tomography (SPET), magnetic resonance imaging (MRI) and electroencephalography (EEG) is described. For SPET studies, technetium-99m ethyl cysteinate dimer (ECD) was injected intravenously while the patient was monitored on video-EEG to document the ictal or interictal state. Imaging was performed using a triple-head gamma camera equipped with a transmission imaging device using a gadolinium-153 source. The images (128x128 pixels, voxel size 3.7x3.7x3.6 mm3) were reconstructed using an iterative algorithm and postfiltered with a Wiener filter. The gold-plated silver electrodes on the patient's scalp were utilized as markers for registration of the ictal and interictal SPET images, as these metallic markers were clearly seen on the transmission images. Fitting of the marker sets was based on a non-iterative least squares method. The interictal SPET image was subtracted from the ictal image after scaling. The T1-weighted MPRAGE MR images with voxel size of 1.0x1.0x1.0 mm3 were obtained with a 1.5-T scanner. For registration of MR and subtraction SPET images, the external marker set of the ictal SPET study was fitted to the surface of the head segmented from MR images. The SPET registration was tested with a phantom experiment. Registration of ictal and interictal SPET in five patient studies resulted in a 2-mm RMS residual of the marker sets. The estimated RMS error of registration in the final result combining locations of the electrodes, subtraction SPET and MR images was 3-5 mm. In conclusion, transmission imaging can be utilized for an accurate and easily implemented registration procedure for ictal and interictal SPET, MRI and EEG.