Tissue segmentation on MR images of the brain by possibilistic clustering on a 3D wavelet representation

An algorithm for the segmentation of a single sequence of three-dimensional magnetic resonance (MR) images into cerebrospinal fluid, gray matter, and white matter classes is proposed. This new method is a possibilistic clustering algorithm using the fuzzy theory as frame and the wavelet coefficients of the voxels as features to be clustered. Fuzzy logic models the uncertainty and imprecision inherent in MR images of the brain, while the wavelet representation allows for both spatial and textural information. The procedure is fast, unsupervised, and totally independent of any statistical assumptions. The method is tested on a phantom image, then applied to normal and Alzheimer's brains, and finally compared with another classic brain tissue segmentation method, affording a relevant classification of voxels into the different tissue classes.     

Interactive segmentation of cerebral gray matter, white matter, and CSF: Photographic and MR images

Digital photography of postmortem brain slices was compared with magnetic resonance imaging (MRI) for morphological analysis of human brain atrophy. In this study, we used two human brains obtained at autopsy: a cognitively defined nondemented control (70-yr-old male) and a demented Alzheimer's disease (AD) subject (82yr-old female). For each of two brains, interactive manual image segmentation was performed by two observers on two image sets: (a) four coronal T1-weighted MR images (5 mm slices); and (b) four digitized photographic images from comparable rostrocaudal levels. Microcomputer image analysis software was used to measure the areas of three segmented cerebral compartments—gray matter (GM), white matter (WM) and CSF—for both image types. Resegmentation error was defined as the absolute difference between the areas derived from two segmentation trials divided by the value from trial 1 and multiplied by 100. This yielded the percent difference between the area measurements from the two trials. We found intea-observer agreement was better (error rates 1–18%) than inter-observer agreement (3–70%) with best agreement for WM and least for CSF, the smallest object class. MRI overestimated GM area relative to digitized photographs in the control but not the AD brain. The results define limitations of manual image segmentations and comparison of MRI with pathologic section photographic images.     

Adult and neonatal human brain: diffusional anisotropy and myelination with diffusion-weighted MR imaging

Diffusional anisotropy of human brain was investigated clinically in six adult volunteers, eight premature neonates, and three infants aged 5-10 months. Diffusion-weighted magnetic resonance imaging was performed with gradient b of 450 sec/mm2. The direction of diffusion-sensitive gradients was changed among x, y, and z axes according to the orientation of neurofibers in white matter. T1- and T2-weighted images also were obtained for evaluation of myelination. Diffusional anisotropy was demonstrated in white matter in the adults. Extensive signal attenuation was observed when gradients were parallel to white matter fibers. Conversely, in neonates, diffusional anisotropy of white matter, in which no myelination was shown on T1- and T2-weighted images, was weak. Diffusional anisotropy was more distinct after brain maturation, as observed in adult white matter. Detection of diffusional anisotropy is useful in evaluating neonatal brain development and various white matter disorders.     

Segmentation of MR brain images into cerebrospinal fluid spaces, white and gray matter

An interactive computer method for quantifying CSF, white matter, and gray matter in magnetic resonance (MR) axial brain scans is presented. A stripping algorithm is used to remove the skull and scalp from each axial section. The images are then filtered to correct for radiofrequency inhomogeneity image artifacts. Late echo images are subtracted from or added to early echo images to enhance fluid/tissue and gray/white tissue contrast, respectively. Thresholds for fluid/tissue and gray/white separation are set interactively. A boundary pixel locking algorithm is used to handle ambiguities due to partial voluming between the fluid and tissue compartments. The MR brain scans from five healthy, young, normal men were obtained using a standard neuroanatomical reference technique. These data were processed and percentages computed for fluid, gray matter and white matter compartments. The gray/white ratios compare favorably with those determined in a published postmortem brain study.     

Infants with perinatal hypoxic ischemia: feasibility of fiber tracking at birth and 3 months

The purpose of this study was to retrospectively investigate the feasibility of fiber tracking at birth and 3 months in infants with hypoxic ischemia to detect disturbances in white matter development. This retrospective study did not require institutional review board approval. All parents gave informed consent. Diffusion-tensor magnetic resonance (MR) images were obtained in full-term newborns: seven with standard MR imaging findings and 10 with perinatal hypoxic ischemia-related abnormalities. Visualization of white matter tracts was investigated by using a volume-tracing and quantification technique. Fiber tracking was useful for studying the neonatal brain. Abnormalities resulted in fiber patterns that were different from the fiber patterns of normal brain tissue. The corona radiata fibers were frequently affected.     

MR image contrast at high field strength.

The T1 of soft tissues increases with magnetic field strength. Some tissue contrast may be diminished on high-field-strength magnetic resonance (MR) images when conventional TRs are used, because of altered T1 effects on the MR signals. This necessitates longer TRs in techniques that use long TRs, which prolongs the examination excessively. Behavior of macroscopic magnetization is governed by the Bloch equations. Therefore, T1 contributions to the MR signal can be modulated by means of both timing intervals and radio-frequency pulses. The analytic solution to the Block equations allowed calculation of white matter/gray matter and gray matter/cerebrospinal fluid contrast in both spin-echo and inversion-recovery (IR) imaging. Rabbit brains (normal and tumor-containing) were then imaged in vivo at 1.5 and 4.7 T. In addition, MR images of a human head were obtained at 4.0 T. Experimental results supported the theoretical predictions that brain contrast on long TR spin-echo or IR images increases with field strength. However, varying the excitation flip angle allowed optimization of the T1 contribution to the MR signals, improving image contrast and/or reducing examination time. Thus, the dependence of T1 on field strength determines the optimum choice of imaging techniques and parameters in a predictable fashion.     

Multispectral analysis of magnetic resonance images

Magnetic resonance (MR) imaging systems produce spatial distribution estimates of proton density, relaxation time, and flow, in a two dimensional matrix form that is analogous to that of the image data obtained from multispectral imaging satellites. Advanced NASA satellite image processing offers sophisticated multispectral analysis of MR images. Spin echo and inversion recovery pulse sequence images were entered in a digital format compatible with satellite images and accurately registered pixel by pixel. Signatures of each tissue class were automatically determined using both supervised and unsupervised classification. Overall tissue classification was obtained in the form of a theme map. In MR images of the brain, for example, the classes included CSF, gray matter, white matter, subcutaneous fat, muscle, and bone. These methods provide an efficient means of identifying subtle relationships in a multi-image MR study.     

Automated classification of multi-spectral MR images using Linear Discriminant Analysis

Magnetic resonance imaging (MRI) is a valuable instrument in medical science owing to its capabilities in soft tissue characterization and 3D visualization. A potential application of MRI in clinical practice is brain parenchyma classification. This work proposes a novel approach called “Unsupervised Linear Discriminant Analysis (ULDA)” to classify and segment the three major tissues, i.e. gray matter (GM), white matter (WM) and cerebral spinal fluid (CSF), from a multi-spectral MR image of the human brain. The ULDA comprises two processes, namely Target Generation Process (TGP) and Linear Discriminant Analysis (LDA) classification. TGP is a fuzzy-set process that generates a set of potential targets from unknown information, and applies these targets to train the optimal division boundary by LDA, such that three tissues GM, WM and CSF are separated. Finally, two sets of images, namely computer-generated phantom images and real MR images are used in the experiments to evaluate the effectiveness of ULDA. Experiment results reveal that UDLA segments a multi-spectral MR image much more effectively than either FMRIB's Automated Segmentation Tool (FAST) or Fuzzy C-means (FC).     

Multiparametric display of spin-echo data from MR studies of brain

A magnetic resonance (MR) image processing technique that uses a single color image for simultaneous presentation of spin-echo information and its application to MR studies of the brain is described. Relaxation rate and proton-density maps were calculated from 160 brain MR studies performed at 1.5 and 1.0 T with standard spin-echo sequences. Maps were fused into single color images, with R1, R2. and proton density coded, respectively, by red, green, and blue. The possibility of standardizing the technique was evaluated. Comparative analysis of color and conventional MR images of white matter disease and brain tumors was performed to assess intra- and interob-server variability. Unequivocal and reproducible chromatic characterization of normal brain structures and a variety of lesions was obtained. Intra-and interobserver analysis showed that color images can be used as a diagnostic tool. The technique may provide a simplified and timesaving approach for interpretation and presentation of brain MR studies.     

SPECT evaluation of simulated white matter lesions: experimental study using a brain phantom.

An experimental study using a brain phantom was performed to evaluate the detectability of simulated white matter lesions on two types of single photon emission computed tomography (SPECT) systems. A ring-type SPECT system with high spatial resolution was able to demonstrate these lesions in the white matter, while a widely used general-purpose single-head rotating gamma camera failed to show them. The detectability of white matter lesions was decreased by photons scattered from the highly radioactive gray matter and the poor spatial resolution of the SPECT system. In order to improve the detectability of white matter lesions it is important to apply scatter correction and improve the spatial resolution of the SPECT system. Brain phantom studies made it easier to interpret white matter lesions that were difficult to interpret on the basis of clinical images alone, and the knowledge acquired from the brain phantom study will contribute to the interpretation of clinical images.     

Regional differences in the proton magnetic resonance relaxation times T1 and T2 within the normal human brain

The proton magnetic resonance (MR) relaxation times T1 and T2 were determined in autopsy specimens from 13 different regions of normal human brains. One hundred and seventy-four tissue samples from 25 brains were examined in a pulsed MR analyzer of 0.25 T and were then also studied histologically. There were regional differences in T1 and T2 within the cerebral gray matter but not within the white matter. These regional differences might reflect the different composition and cytoarchitectonic structure of the cortical regions and should be taken into consideration in the interpretation of cortical lesions on MR images.     

Brain CT and MR findings in hyperphenylalaninemia due to dihydropteridine reductase deficiency (variant of phenylketonuria)

Two siblings with malignant hyperphenylalaninemia were examined by magnetic resonance (MR) imaging and CT of the brain. Both techniques demonstrated diffuse cerebral atrophy and cystic loss of parenchyma with surrounding white matter changes. T2-weighted MR images demonstrated the white matter changes better than CT. However, MR images gave no definite indication of the presence of calcification, and CT demonstrated the characteristic calcifications in the basal ganglia and subcortical region bilaterally. Both MR and CT are complementary in the evaluation of this disease.     

Brain surface cortical sulcal lengths: quantification with three-dimensional MR imaging

The repeatability and accuracy of brain surface cortical sulcal length measurements obtained with three-dimensional (3D) reconstructions of volumetric, gradient-echo magnetic resonance (MR) images were tested. The brains of eight healthy adult volunteers and one cadaver were imaged in both the coronal and sagittal planes to yield a set of 128 1.5-2.0-mm-thick contiguous sections. 3D reconstructions of the brain cerebral cortical surfaces were obtained with computer software. Location and distance measurements of surface sulci were repeated on each reconstructed image. The same structures in the cadaver brain were independently measured with a 3D electromagnetic digitizer to validate the results of the 3D MR imaging method. All measurements from reconstructed images had high repeatability, and there were no statistically significant differences between measurement trials. The accuracy of measurements with 3D MR imaging was also good; the mean difference between digitizer and 3D MR measurements for sulcal lengths was 0.81 cm (average, 5.45-12.9 cm).     

Anatomic dissection tractography: a new method for precise MR localization of white matter tracts

BACKGROUND AND PURPOSE: Several white matter tracts in the brain cannot be identified on MR studies because they are indistinguishable from the surrounding white matter. We sought to develop a method to precisely localize white matter tracts by correlating anatomic dissections with corresponding MR images. METHODS: MR imaging was used to guide anatomic dissection of the uncinate fasciculus. Formalin-preserved brains were imaged before and after several stages of dissection. Progressive dissection was guided by using volume-rendered and cross-sectional images of the dissected specimens. To precisely define the location of a tract, its surface was traced on the corresponding three-dimensional MR image of the dissected specimen. MR images of the dissected and intact specimens were coregistered to allow the tracings to be projected onto multiplanar reformatted images of the intact specimen. RESULTS: The uncinate fasciculus in the anterior temporal lobe and external and extreme capsules was dissected without destroying adjacent structures. Coregistration of the MR images from intact and dissected specimens permitted precise MR identification of the surface of this tract. These methods were successful for two additional tracts. (The dissected anatomy, MR anatomy, and clinical examples of the three tracts are described in a companion article.) CONCLUSION: MR-assisted anatomic dissection permits limited removal of brain tissue so that important anatomic and surgical relationships can be demonstrated on correlated MR studies. This method can be applied to other white matter tracts that are indistinguishable on MR studies and to situations in which anatomic validation of normal and abnormal diffusion tractographic studies is needed.     

Precise and accurate measurement of proton T1 in human brain in vivo: Validation and preliminary clinical application

Precise and accurate inversion-recovery (PAIR) magnetic resonance (MR) measurements of T1 were obtained in eight brain regions and cerebrospinal fluid of 26 healthy volunteers. Accuracy of the technique was assessed by measuring T1 in small fluid volumes with the PAIR technique and with two independent spectroscopic techniques. The mean difference between T1 measured with PAIR and with the two spectroscopic techniques was 3.1% ± 1.3. The precision (reproducibility) of measurements with the PAIR technique was excellent. The coefficient of variation (CV) across 16 measurements in a head phantom was 2.0%, compared with a CV of 2.7% across 45 separate measurements in a single subject. The within-subject CV was 1.8% ± 0.6 in white matter and 1.4% ± 1.0 in basal ganglia. The between-subject CV in 26 healthy volunteers was 3.6% ± 0.6 in white matter and 4.1% ± 1.9 in basal ganglia. Comparison between a patient with an active recurrent brain tumor and an agematched patient with an inactive brain tumor showed that T1 was significantly elevated throughout the brain of the active-tumor patient, especially in white matter tracts, even though no tumor or edema was detected in the white matter on standard MR images. Comparisons between five brain tumor patients and four healthy volunteers of similar age showed that T1 was significantly and substantially elevated throughout the white matter tracts and in the caudate nucleus, putamen, and thalamus. These results are consistent with the hypothesis that white matter tracts are selectively vulnerable to edema and that T1 increases in white matter are a sensitive indicator of patient status or tumor aggressiveness.     

Discriminating between silent cerebral infarction and deep white matter hyperintensity using combinations of three types of magnetic resonance images: a multicenter observer performance study

INTRODUCTION: We attempted to determine the most appropriate combination of magnetic resonance (MR) images that can accurately detect and discriminate between asymptomatic infarction and deep white matter hyperintensity (DWMH); these lesions have different clinical implications and are occasionally confused. MATERIALS AND METHODS: We performed an observer performance analysis using cerebral MR images of 45 individuals with or without asymptomatic small white matter infarction and/or mild DWMH who participated in a physical checkup program at four institutions. Six observers interpreted whether infarction and/or DWMH existed in combinations of two or three image types of the T1-weighted images (T1WI), T2-weighted images (T2WI), and fluid-attenuated inversion recovery (FLAIR) images. The observers' performance was evaluated with a receiver operating characteristic (ROC) analysis. RESULTS: The averaged area under the ROC curve (Az) for detecting a infarction was significantly larger in the combination of all the three image types (0.95) than that in any combinations of the two image types (T1WI and FLAIR images, 0.87; T2WI and FLAIR images, 0.85; T1WI and T2WI, 0.86). The Az for detecting DWMH was significantly smaller in the combination of T1WI and T2WI (0.79) than that in other image combinations (T1WI and FLAIR, 0.89; T2WI and FLAIR, 0.91; T1WI, T2WI, and FLAIR, 0.90). CONCLUSION: The combination of T1WI, T2WI, and FLAIR images is required to accurately detect both small white matter infarction and mild DWMH.     

Evaluation of three-dimensional fast spoiled gradient recalled acquisition in the steady state (FSPGR) using ultra magnetic field 3-Tesla MRI for optimal pulse sequences of T1-weighted imaging

The advantage of the higher signal-to-noise ratio (SNR) of 3-Tesla magnetic resonance imaging (3TMRI) contributes to the improvement of spatial and temporal resolution. However, T1-weighted images of the brain obtained by the spin-echo (SE) method at 3T MR are not satisfactory for clinical use because of radiofrequency (RF) field inhomogeneity and prolongation of the longitudinal relaxation time (T1) of most tissues. We evaluated optimal pulse sequences to obtain adequate T1 contrast, high gray matter/white matter contrast, and suitable postcontrast T1-weighted images using the three-dimentional (3D) fast spoiled gradient recalled acquisition in the steady state (FSPGR) method instead of the SE method. For the optimization of T1 contrast, the Ernst angle of the optimal flip angle (FA) was obtained from the T1 value of cerebral white matter with the shortest TR and TE. Then the most appropriate FA, showing the maximum contrast-to-noise ratio (CNR) and SNR, was obtained by changing the FA every 5 degrees at about the level of the Ernst angle. Image uniformity was evaluated by a phantom showing similar T1 and T2 values of cerebral white matter. In order to evaluate the effect of the contrast enhancement, signal intensity was compared by the same method using a phantom filled with various dilutions of contrast media. Moreover, clinical studies using full (0.1 mmol/kg) and half (0.05 mmol/kg) doses of Gd-DTPA were carried out with the most appropriate parameters of the 3D-FSPGR method. These studies indicated that the optimal pulse sequences for obtaining an adequate T1-weighted image of the brain using 3D-FSPGR are 9/2 msec (TR/TE) and 13 degrees (FA).     

Sensitivity and reproducibility of a new fast 3D segmentation technique for clinical MR-based brain volumetry in multiple sclerosis

Fast, reliable and easy-to-use methods to quantify brain atrophy are of increasing importance in clinical studies on neuro-degenerative diseases. Here, ILAB 4, a new volumetry software that uses a fast semi-automated 3D segmentation of thin-slice T1-weighted 3D MR images based on a modified watershed transform and an automatic histogram analysis was evaluated. It provides the cerebral volumes: whole brain, white matter, gray matter and intracranial cavity. Inter- and intra-rater reliability and scan-rescan reproducibility were excellent in measuring whole brain volumes (coefficients of variation below 0.5%) of volunteers and patients. However, gray and white matter volumes were more susceptible to image quality. High accuracy of the absolute volume results (+/-5 ml) were shown by phantom and preparation measurements. Analysis times were 6 min for processing of 128 slices. The proposed technique is reliable and highly suitable for quantitative studies of brain atrophy, e.g., in multiple sclerosis.     

Comparison of 31P MRS and 1H MRI at 1.5 and 2.0 T

The goals of this study were to compare 31P magnetic resonance spectroscopy (MRS) and 1H magnetic resonance imaging (MRI) of human subjects and phantoms at 1.5 and 2.0 T. The 31P signal-to-noise (S/N) ratios in phantom standards and in localized volumes in human brain and liver were compared at 1.5 and 2.0 T. In addition, T1 values for 31P resonances in human brain, 31P linewidths of metabolites in human brain and liver, 1H S/N in a phantom standard, and MR image quality in human head and body were compared at the two field strengths. The results of our study showed that at the higher strength field, (1) in vivo 31P MRS studies benefited from up to 32% improvement in S/N; (2) in vivo 31P MRS studies also benefited from increased spectral dispersion; (3) the quality of MR head images remained comparable; and (4) body images showed some decrease in image quality due to increased chemical shift, and flow and motion artifacts.     

High-resolution sodium imaging of human brain at 7 T

The feasibility of high-resolution sodium magnetic resonance imaging on human brain at 7 T was demonstrated in this study. A three-dimensional anisotropic resolution data acquisition was used to address the challenge of low signal-to-noise ratio associated with high resolution. Ultrashort echo-time sequence was used for the anisotropic data acquisition. Phantoms and healthy human brains were studied on a whole-body 7-T magnetic resonance imaging scanner. Sodium images were obtained at two high nominal in-plane resolutions (1.72 and 0.86 mm) at a slice thickness of 4 mm. Signal-to-noise ratio in the brain image (cerebrospinal fluid) was measured as 14.4 and 6.8 at the two high resolutions, respectively. The actual in-plane resolution was measured as 2.9 and 1.6 mm, 69-86% larger than their nominal values. The quantification of sodium concentration on the phantom and brain images enabled better accuracy at the high nominal resolutions than at the low nominal resolution of 3.44 mm (measured resolution 5.5 mm) due to the improvement of in-plane resolution.