Joseph V. Hajnal

Automatic anatomical brain MRI segmentation combining label propagation and decision fusion.

Regions in three-dimensional magnetic resonance (MR) brain images can be classified using protocols for manually segmenting and labeling structures. For large cohorts, time and expertise requirements make this approach impractical. To achieve automation, an individual segmentation can be propagated to another individual using an anatomical correspondence estimate relating the atlas image to the target image. The accuracy of the resulting target labeling has been limited but can potentially be improved by combining multiple segmentations using decision fusion. We studied segmentation propagation and decision fusion on 30 normal brain MR images, which had been manually segmented into 67 structures. Correspondence estimates were established by nonrigid registration using free-form deformations. Both direct label propagation and an indirect approach were tested. Individual propagations showed an average similarity index (SI) of 0.754+/-0.016 against manual segmentations. Decision fusion using 29 input segmentations increased SI to 0.836+/-0.009. For indirect propagation of a single source via 27 intermediate images, SI was 0.779+/-0.013. We also studied the effect of the decision fusion procedure using a numerical simulation with synthetic input data. The results helped to formulate a model that predicts the quality improvement of fused brain segmentations based on the number of individual propagated segmentations combined. We demonstrate a practicable procedure that exceeds the accuracy of previous automatic methods and can compete with manual delineations.

Multi-atlas based segmentation of brain images: Atlas selection and its effect on accuracy

Quantitative research in neuroimaging often relies on anatomical segmentation of human brain MR images. Recent multi-atlas based approaches provide highly accurate structural segmentations of the brain by propagating manual delineations from multiple atlases in a database to a query subject and combining them. The atlas databases which can be used for these purposes are growing steadily. We present a framework to address the consequent problems of scale in multi-atlas segmentation. We show that selecting a custom subset of atlases for each query subject provides more accurate subcortical segmentations than those given by non-selective combination of random atlas subsets. Using a database of 275 atlases, we tested an image-based similarity criterion as well as a demographic criterion (age) in a leave-one-out cross-validation study. Using a custom ranking of the database for each subject, we combined a varying number n of atlases from the top of the ranked list. The resulting segmentations were compared with manual reference segmentations using Dice overlap. Image-based selection provided better segmentations than random subsets (mean Dice overlap 0.854 vs. 0.811 for the estimated optimal subset size, n=20). Age-based selection resulted in a similar marked improvement. We conclude that selecting atlases from large databases for atlas-based brain image segmentation improves the accuracy of the segmentations achieved. We show that image similarity is a suitable selection criterion and give results based on selecting atlases by age that demonstrate the value of meta-information for selection.

Natural History of Brain Lesions in Extremely Preterm Infants Studied With Serial Magnetic Resonance Imaging From Birth and Neurodevelopmental Assessment

OBJECTIVES. The aim was to survey the range of cerebral injury and abnormalities of cerebral development in infants born between 23 and 30 weeks’ gestation using serial MRI scans of the brain from birth, and to correlate those findings with neurodevelopmental outcome after 18 months corrected age. METHODS. Between January 1997 and November 2000, consecutive infants born at <30 weeks’ gestational age underwent serial MRI brain scans from birth until term-equivalent age. Infants were monitored after 18 months of age, corrected for prematurity, with the Griffiths Mental Development Scales and neurologic assessment. RESULTS. A total of 327 MRI scans were obtained from 119 surviving infants born at 23 to 30 weeks of gestation. Four infants had major destructive brain lesions, and tissue loss was seen at term for the 2 survivors. Fifty-one infants had early hemorrhage; 50% of infants with term scans after intraventricular hemorrhage had ventricular dilation. Twenty-six infants had punctate white matter lesions on early scans; these persisted for 33% of infants assessed at term. Early scans showed cerebellar hemorrhagic lesions for 8 infants and basal ganglia abnormalities for 17. At term, 53% of infants without previous hemorrhage had ventricular dilation and 80% of infants had diffuse excessive high signal intensity within the white matter on T2-weighted scans. Complete follow-up data were available for 66% of infants. Adverse outcomes were associated with major destructive lesions, diffuse excessive high signal intensity within the white matter, cerebellar hemorrhage, and ventricular dilation after intraventricular hemorrhage but not with punctate white matter lesions, hemorrhage, or ventricular dilation without intraventricular hemorrhage. CONCLUSIONS. Diffuse white matter abnormalities and post–hemorrhagic ventricular dilation are common at term and seem to correlate with reduced developmental quotients. Early lesions, except for cerebellar hemorrhage and major destructive lesions, do not show clear relationships with outcomes.

Use of multicoil arrays for separation of signal from multiple slices simultaneously excited

Increased acquisition efficiency has been achieved by exciting several slices simultaneously. The mixed data were unfolded to produce separate slices using the spatial encoding information inherent in a multicoil receiver system. Each coil yields a linear combination of signals from all excited slices weighted by the sensitivity of each coil. A matrix inversion provides a solution to unfold these images. J. Magn. Reson. Imaging 2001;13:313–317.

A registration and interpolation procedure for subvoxel matching of serially acquired MR images

Objective Methods for automatically registering and reslicing MR images using an interpolation function that matches the structure of the image data are described. Materials and Methods Phantom and human brain images were matched by rigid body rotations and translations in two and three dimensions using a least-squares optimization procedure. Subvoxel image shifts were produced with linear or sine interpolation. Results The use of sine interpolation ensured that the repositioned images were faithful to the original data and enabled quantitative intensity comparisons to be made. In humans, image segmentation was vital to avoid extraneous soft tissue changes producing systematic errors in registration. Conclusion The sinc-based interpolation technique enabled serially acquired MR images to be positionally matched to subvoxel accuracy so that small changes in the brain could be distinguished from effects due to misregistration. Index Terms Image registration—Brain—Phantom—Magnetic resonance imaging.

Diffusion tensor imaging with tract-based spatial statistics reveals local white matter abnormalities in preterm infants

Infants born preterm have a high incidence of neurodevelopmental impairment in later childhood, often associated with poorly defined cerebral white matter abnormalities. Diffusion tensor imaging quantifies the diffusion of water within tissues and can assess microstructural abnormalities in the developing preterm brain. Tract-based spatial statistics (TBSS) is an automated observer-independent method of aligning fractional anisotropy (FA) images from multiple subjects to allow groupwise comparisons of diffusion tensor imaging data. We applied TBSS to test the hypothesis that preterm infants have reduced fractional anisotropy in specific regions of white matter compared to term-born controls. We studied 26 preterm infants with no evidence of focal lesions on conventional magnetic resonance imaging (MRI) at term equivalent age and 6 healthy term-born control infants. We found that the centrum semiovale, frontal white matter and the genu of the corpus callosum showed significantly lower FA in the preterm group. Infants born at less than or equal to 28 weeks gestational age (n=11) displayed additional reductions in FA in the external capsule, the posterior aspect of the posterior limb of the internal capsule and the isthmus and middle portion of the body of the corpus callosum. This study demonstrates that TBSS provides an observer-independent method of identifying white matter abnormalities in the preterm brain at term equivalent age in the absence of focal lesions.

Angiotensin-converting-enzyme gene insertion/deletion polymorphism and response to physical training.

BACKGROUND The function of local renin-angiotensin systems in skeletal muscle and adipose tissue remains largely unknown. A polymorphism of the human angiotensin converting enzyme (ACE) gene has been identified in which the insertion (I) rather than deletion (D) allele is associated with lower ACE activity in body tissues and increased response to some aspects of physical training. We studied the association between the ACE gene insertion or deletion polymorphism and changes in body composition related to an intensive exercise programme, to investigate the metabolic effects of local human renin-angiotensin systems. METHODS We used three independent methods (bioimpedance, multiple skinfold-thickness assessment of whole-body composition, magnetic resonance imaging of the mid-thigh) to study changes in body composition in young male army recruits over 10 weeks of intensive physical training. FINDINGS Participants with the II genotype had a greater anabolic response than those with one or more D alleles for fat mass (0.55 vs -0.20 kg, p=0.04 by bioimpedance) and non-fat mass (1.31 vs -0.15 kg, p=0.01 by bioimpedance). Changes in body morphology with training measured by the other methods were also dependent on genotype. INTERPRETATION II genotype, as a marker of low ACE activity in body tissues, may conserve a positive energy balance during rigorous training, which suggests enhanced metabolic efficiency. This finding may explain some of the survival and functional benefits of therapy with ACE inhibitors.

Axial and Radial Diffusivity in Preterm Infants Who Have Diffuse White Matter Changes on Magnetic Resonance Imaging at Term-Equivalent Age

Objective. Diffuse excessive high signal intensity (DEHSI) is observed in the majority of preterm infants at term-equivalent age on conventional MRI, and diffusion-weighted imaging has shown that apparent diffusion coefficient values are elevated in the white matter (WM) in DEHSI. Our aim was to obtain diffusion tensor imaging on preterm infants at term-equivalent age and term control infants to test the hypothesis that radial diffusivity was significantly different in the WM in preterm infants with DEHSI compared with both preterm infants with normal-appearing WM on conventional MRI and term control infants. Methods. Diffusion tensor imaging was obtained on 38 preterm infants at term-equivalent age and 8 term control infants. Values for axial (λ1) and radial [(λ2 + λ3)/2] diffusivity were calculated in regions of interest positioned in the central WM at the level of the centrum semiovale, frontal WM, posterior periventricular WM, occipital WM, anterior and posterior portions of the posterior limb of the internal capsule, and the genu and splenium of the corpus callosum. Results. Radial diffusivity was elevated significantly in the posterior portion of the posterior limb of the internal capsule and the splenium of the corpus callosum, and both axial and radial diffusivity were elevated significantly in the WM at the level of the centrum semiovale, the frontal WM, the periventricular WM, and the occipital WM in preterm infants with DEHSI compared with preterm infants with normal-appearing WM and term control infants. There was no significant difference between term control infants and preterm infants with normal-appearing WM in any region studied. Conclusions. These findings suggest that DEHSI represents an oligodendrocyte and/or axonal abnormality that is widespread throughout the cerebral WM.

Specific relations between neurodevelopmental abilities and white matter microstructure in children born preterm

Survivors of preterm birth have a high incidence of neurodevelopmental impairment which is not explained by currently understood brain abnormalities. The aim of this study was to test the hypothesis that the neurodevelopmental abilities of 2-year-old children who were born preterm and who had no evidence of focal abnormality on conventional MR imaging were consistently linearly related to specific local changes in white matter microstructure. We studied 33 children, born at a median (range) gestational age of 28(+5) (24(+4)-32(+1)) weeks. The children were recruited as infants from the Neonatal Intensive Care Unit at Queen Charlottes and Hammersmith Hospital in the early neonatal period and imaged at a median corrected age of 25.5 (24-27) months. The children underwent diffusion tensor imaging to measure fractional anisotropy (FA) as a measure of tissue microstructure, and neurodevelopmental assessment using the Griffiths Mental Development Scales [giving an overall developmental quotient (DQ) and sub-quotients scores for motor, personal-social, hearing-language, eye-hand coordination and performance scales] at 2 years corrected age. Tract-based spatial statistics with linear regression analysis of voxel-wise cross-subject statistics were used to assess the relationship between FA and DQ/sub-quotient scores and results confirmed by reduced major axis regression of regions with significant correlations. We found that DQ was linearly related to FA values in parts of the corpus callosum; performance sub-scores to FA values in the corpus callosum and right cingulum; and eye-hand coordination sub-scores to FA values in the cingulum, fornix, anterior commissure, corpus callosum and right uncinate fasciculus. This study shows that specific neurodevelopmental impairments in infants born preterm are precisely related to microstructural abnormalities in particular regions of cerebral white matter which are consistent between individuals. FA may aid prognostication and provide a biomarker for therapeutic or mechanistic studies of preterm brain injury.

Diffeomorphic registration using b-splines

In this paper we propose a diffeomorphic non-rigid registration algorithm based on free-form deformations (FFDs) which are modelled by B-splines. In contrast to existing non-rigid registration methods based on FFDs the proposed diffeomorphic non-rigid registration algorithm based on free-form deformations (FFDs) which are modelled by B-splines. To construct a diffeomorphic transformation we compose a sequence of free-form deformations while ensuring that individual FFDs are one-to-one transformations. We have evaluated the algorithm on 20 normal brain MR images which have been manually segmented into 67 anatomical structures. Using the agreement between manual segmentation and segmentation propagation as a measure of registration quality we have compared the algorithm to an existing FFD registration algorithm and a modified FFD registration algorithm which penalises non-diffeomorphic transformations. The results show that the proposed algorithm generates diffeomorphic transformations while providing similar levels of performance as the existing FFD registration algorithm in terms of registration accuracy.