RATIONALE AND OBJECTIVES: To examine a statistical validation method based on the spatial overlap between two sets of segmentations of the same anatomy. MATERIALS AND METHODS: The Dice similarity coefficient (DSC) was used as a statistical validation metric to evaluate the performance of both the reproducibility of manual segmentations and the spatial overlap accuracy of automated probabilistic fractional segmentation of MR images, illustrated on two clinical examples. Example 1: 10 consecutive cases of prostate brachytherapy patients underwent both preoperative 1.5T and intraoperative 0.5T MR imaging. For each case, 5 repeated manual segmentations of the prostate peripheral zone were performed separately on preoperative and on intraoperative images. Example 2: A semi-automated probabilistic fractional segmentation algorithm was applied to MR imaging of 9 cases with 3 types of brain tumors. DSC values were computed and logit-transformed values were compared in the mean with the analysis of variance (ANOVA). RESULTS: Example 1: The mean DSCs of 0.883 (range, 0.876-0.893) with 1.5T preoperative MRI and 0.838 (range, 0.819-0.852) with 0.5T intraoperative MRI (P < .001) were within and at the margin of the range of good reproducibility, respectively. Example 2: Wide ranges of DSC were observed in brain tumor segmentations: Meningiomas (0.519-0.893), astrocytomas (0.487-0.972), and other mixed gliomas (0.490-0.899). CONCLUSION: The DSC value is a simple and useful summary measure of spatial overlap, which can be applied to studies of reproducibility and accuracy in image segmentation. We observed generally satisfactory but variable validation results in two clinical applications. This metric may be adapted for similar validation tasks.
The injury of Phineas Gage has fueled research on and fascination with the localization of cerebral functions in the past century and a half. Most physicians and anatomists believed that Gage sustained a largely bilateral injury to the frontal lobes. However, previous studies seem to have overlooked a few less obvious, but essential details. This has led us to reanalyze the injury using three-dimensional reconstruction and quantitative computer-aided techniques and to propose a new biomechanical model, in order to determine the location and extent of the injury and explain Gage's improbable survival. Unlike previous studies on this subject, our findings are based on computer-generated three-dimensional reconstructions of a thin-slice computed tomography scan (CAT) of Phineas Gage's skull. The results of our image analysis were corroborated with the clinical findings, thoroughly recorded by Dr. Harlow in 1848, as well as with a systematic examination of the original skull specimen. Our results show that the cerebral injury was limited to the left frontal lobe, did not extend to the contralateral side, did not affect the ventricular system, and did not involve vital intracranial vascular structures. Although modern neuroscience has perhaps outgrown the speculations prompted by this famous case, it is still a living part of the medical folklore and education. Setting the record straight based on clinical reasoning, observation of the physical evidence, and sound quantitative computational methods is more than mere minutia and of interest for the broad medical community.
The validity of brain tumour segmentation is an important issue in image processing because it has a direct impact on surgical planning. We examined the segmentation accuracy based on three two-sample validation metrics against the estimated composite latent gold standard, which was derived from several experts' manual segmentations by an EM algorithm. The distribution functions of the tumour and control pixel data were parametrically assumed to be a mixture of two beta distributions with different shape parameters. We estimated the corresponding receiver operating characteristic curve, Dice similarity coefficient, and mutual information, over all possible decision thresholds. Based on each validation metric, an optimal threshold was then computed via maximization. We illustrated these methods on MR imaging data from nine brain tumour cases of three different tumour types, each consisting of a large number of pixels. The automated segmentation yielded satisfactory accuracy with varied optimal thresholds. The performances of these validation metrics were also investigated via Monte Carlo simulation. Extensions of incorporating spatial correlation structures using a Markov random field model were considered.
In the longitudinal study of multiple sclerosis (MS) lesions, varying position of the patient inside the MRI scanner is one of the major sources of assessment errors. We propose to use analytical indices that are invariant to spatial orientation to describe the lesions, rather than focus on patient repositioning or image realignment. Studies were made on simulated lesions systematically rotated, from in vitro MS lesions scanned on different days, and from in vivo MS lesions from a patient that was scanned five times the same day with short intervals of time between scans. Each of the lesions' 3D surfaces was approximated using spherical harmonics, from which indices that are invariant to space rotation were derived. From these indices, an accurate and highly reproducible volume estimate can be derived, which is superior to the common approach of 2D slice stacking. The results indicate that the suggested approach is useful in reducing part of the errors that affect the analysis of changes of MS lesions during follow-up studies. In conclusion, our proposed method circumvents the need for precise patient repositioning and can be advantageous in MRI longitudinal studies of MS patients.
A high prevalence of abnormal cavum septi pellucidi (CSP) in schizophrenia may reflect neurodevelopmental abnormalities in midline structures of the brain. The relationship, however, between abnormal CSP and clinical symptoms, and with abnormalities in other limbic structures remains unclear, as does the question of whether a similar abnormality is present in affective psychosis. Seventy-four patients at their first hospitalization, 33 with schizophrenia and 41 with affective (mainly manic) psychosis, and 56 healthy control subjects underwent high-spatial-resolution magnetic resonance imaging (MRI). CSP on six slices or more on 0.9375-mm resampled coronal images was categorized as abnormal. The prevalence of abnormal CSP in both schizophrenic patients (26.1%) and affective psychosis patients (18.2%) was significantly higher than was observed in control subjects (8.2%). In schizophrenic patients only, larger CSP was significantly associated with more severe thinking disturbance and smaller left parahippocampal gyrus gray matter volumes. While the relationships between CSP ratings and clinical symptoms did not significantly differ between the two psychosis groups as assessed by the comparison of regression slopes, the association with limbic volumes appeared to be specific to schizophrenic patients. These results suggest that psychosis associated with schizophrenia and affective disorder share, at least to some extent, neurodevelopmental abnormalities involving midline structures and associated psychopathological consequences. However, the association between abnormal CSP and limbic systems may be more specific to schizophrenia.
In this paper we present a framework for unsupervised segmentation of white matter fiber traces obtained from diffusion weighted MRI data. Fiber traces are compared pairwise to create a weighted undirected graph which is partitioned into coherent sets using the normalized cut (N cut) criterion. A simple and yet effective method for pairwise comparison of fiber traces is presented which in combination with the N cut criterion is shown to produce plausible segmentations of both synthetic and real fiber trace data. Segmentations are visualized as colored stream-tubes or transformed to a segmentation of voxel space, revealing structures in a way that looks promising for future explorative studies of diffusion weighted MRI data.
Changes in mean magnetic resonance imaging (MRI)-derived measurements between patient groups are often used to determine outcomes in therapeutic trials and other longitudinal studies of multiple sclerosis (MS). However, in day-to-day clinical practice the changes within individual patients may also be of interest In this paper, we estimated the measurement error of an automated brain tissue quantification algorithm and determined the thresholds for statistically significant change of MRI-derived T2 lesion volume and brain atrophy in individual patients. Twenty patients with MS were scanned twice within 30 min. Brain tissue volumes were measured using the computer algorithm. Brain atrophy was estimated by calculation of brain parenchymal fraction. The threshold of change between repeated scans that represented statistically significant change beyond measurement error with 95% certainty was 0.65 mL for T2 lesion burden and 0.0056 for brain parenchymal fraction. Changes in lesion burden and brain atrophy below these thresholds can be safely (with 95% certainty) explained by measurement variability alone. These values provide clinical neurologists with a useful reference to interpret MRI-derived measures in individual patients.
A novel method for the visual and quantitative analysis of the geometrical relationship between the vascular architecture of the brain and white matter pathology is presented. The cerebro vascular system is implicated in the pathogenesis of many diseases of the cerebral white matter, for example, stroke, microcerebrovascular disease, and multiple sclerosis (MS). In our work, white matter lesions and vessels are depicted using magnetic resonance imaging (MRI) and extracted using image analysis techniques. We focus on measuring distance relationships between white matter lesions and vessels, and distribution of lesions with respect to vessel caliber. Vascular distance maps are generated by computing for each voxel the Euclidean distance to the closest vessel. Analogously, radius maps assign the radius of the closest vessel to each voxel in the image volume. The distance and radius maps are used to analyze the distribution of lesions with respect to the vessels' locations and their calibers. The method was applied to three MS patients to demonstrate its functionality and feasibility. Preliminary findings indicate that larger MS lesions tend to be farther from detected vessels and that the caliber of the vessels nearest to larger lesions tends to be smaller, suggesting a possible role of relative hypoperfusion or hypoxia in lesion formation.
Characterizing the performance of image segmentation approaches has been a persistent challenge. Performance analysis is important since segmentation algorithms often have limited accuracy and precision. Interactive drawing of the desired segmentation by human raters has often been the only acceptable approach, and yet suffers from intra-rater and inter-rater variability. Automated algorithms have been sought in order to remove the variability introduced by raters, but such algorithms must be assessed to ensure they are suitable for the task. The performance of raters (human or algorithmic) generating segmentations of medical images has been difficult to quantify because of the difficulty of obtaining or estimating a known true segmentation for clinical data. Although physical and digital phantoms can be constructed for which ground truth is known or readily estimated, such phantoms do not fully reflect clinical images due to the difficulty of constructing phantoms which reproduce the full range of imaging characteristics and normal and pathological anatomical variability observed in clinical data. Comparison to a collection of segmentations by raters is an attractive alternative since it can be carried out directly on the relevant clinical imaging data. However, the most appropriate measure or set of measures with which to compare such segmentations has not been clarified and several measures are used in practice. We present here an expectation-maximization algorithm for simultaneous truth and performance level estimation (STAPLE). The algorithm considers a collection of segmentations and computes a probabilistic estimate of the true segmentation and a measure of the performance level represented by each segmentation. The source of each segmentation in the collection may be an appropriately trained human rater or raters, or may be an automated segmentation algorithm. The probabilistic estimate of the true segmentation is formed by estimating an optimal combination of the segmentations, weighting each segmentation depending upon the estimated performance level, and incorporating a prior model for the spatial distribution of structures being segmented as well as spatial homogeneity constraints. STAPLE is straightforward to apply to clinical imaging data, it readily enables assessment of the performance of an automated image segmentation algorithm, and enables direct comparison of human rater and algorithm performance.
In this paper we present a new algorithm for 3D medical image segmentation. The algorithm is versatile, fast, relatively simple to implement, and semi-automatic. It is based on minimizing a global energy defined from a learned non-parametric estimation of the statistics of the region to be segmented. Implementation details are discussed and source code is freely available as part of the 3D Slicer project. In addition, a new unified set of validation metrics is proposed. Results on artificial and real MRI images show that the algorithm performs well on large brain structures both in terms of accuracy and robustness to noise.
Hemisphere asymmetry was explored in normal healthy subjects and in patients with schizophrenia using a novel voxel-based tensor analysis applied to fractional anisotropy (FA) of the diffusion tensor. Our voxel-based approach, which requires precise spatial normalization to remove the misalignment of fiber tracts, includes generating a symmetrical group average template of the diffusion tensor by applying nonlinear elastic warping of the demons algorithm. We then normalized all 32 diffusion tensor MRIs from healthy subjects and 23 from schizophrenic subjects to the symmetrical average template. For each brain, six channels of tensor component images and one T2-weighted image were used for registration to match tensor orientation and shape between images. A statistical evaluation of white matter asymmetry was then conducted on the normalized FA images and their flipped images. In controls, we found left-higher-than-right anisotropic asymmetry in the anterior part of the corpus callosum, cingulum bundle, the optic radiation, and the superior cerebellar peduncle, and right-higher-than-left anisotropic asymmetry in the anterior limb of the internal capsule and the anterior limb's prefrontal regions, in the uncinate fasciculus, and in the superior longitudinal fasciculus. In patients, the asymmetry was lower, although still present, in the cingulum bundle and the anterior corpus callosum, and not found in the anterior limb of the internal capsule, the uncinate fasciculus, and the superior cerebellar peduncle compared to healthy subjects. These findings of anisotropic asymmetry pattern differences between healthy controls and patients with schizophrenia are likely related to neurodevelopmental abnormalities in schizophrenia.
A new definition of affine invariant medial axis of planar closed curves is introduced. A point belongs to the affine medial axis if and only if it is equidistant from at least two points of the curve, with the distance being a minimum and given by the areas between the curve and its corresponding chords. The medial axis is robust, eliminating the need for curve denoising. In a dynamical interpretation of this affine medial axis, the medial axis points are the affine shock positions of the affine erosion of the curve. We propose a simple method to compute the medial axis and give examples. We also demonstrate how to use this method to detect affine skew symmetry in real images.
OBJECTIVE: This study was undertaken to develop and test a 3-dimensional (3D) color thickness mapping technique on levator ani imaged with magnetic resonance imaging (MRI). METHODS: Supine MRI datasets from 30 women were studied: 10 asymptomatic, 10 with urodynamic stress incontinence, and 10 with pelvic organ prolapse. Levators were manually outlined, and thickness mapping applied. Three-dimensional models were colored topographically, reflecting levator thickness. Thickness and occurrences of absent levator substance (gaps) were compared across the 3 groups, using nonparametric statistical tests. RESULTS: Color thickness mapping was successful in all subjects. There were statistically significant differences in thickness and gap percentages among the 3 groups of women, with thicker, bulkier levators in asymptomatic women, compared with women with prolapse or urodynamic stress incontinence. CONCLUSION: Color thickness mapping is feasible. It may be used to compare levators in symptomatic and asymptomatic women, to study relationships between levator thickness and pelvic floor dysfunction. This technique can be used in larger studies for hypothesis testing.
We introduce a method for combining fiber tracking from diffusion-tensor (DT) imaging with cortical gray matter parcellation from structural high-spatial-resolution 3D spoiled gradient-recalled acquisition in the steady state images. We applied this method to a tumor case to determine the impact of the tumor on white matter architecture. We conclude that this new method for combining structural and DT imaging data is useful for understanding cortical connectivity and the localization of fiber tracts and their relationship with cortical anatomy and brain abnormalities.
This paper presents extensions which improve the performance of the shape-based deformable active contour model presented earlier in [IEEE Conf. Comput. Vision Pattern Recog. 1 (2001) 463] for medical image segmentation. In contrast to that previous work, the segmentation framework that we present in this paper allows multiple shapes to be segmented simultaneously in a seamless fashion. To achieve this, multiple signed distance functions are employed as the implicit representations of the multiple shape classes within the image. A parametric model for this new representation is derived by applying principal component analysis to the collection of these multiple signed distance functions. By deriving a parametric model in this manner, we obtain a coupling between the multiple shapes within the image and hence effectively capture the co-variations among the different shapes. The parameters of the multi-shape model are then calculated to minimize a single mutual information-based cost criterion for image segmentation. The use of a single cost criterion further enhances the coupling between the multiple shapes as the deformation of any given shape depends, at all times, upon every other shape, regardless of their proximity. We found that this resulting algorithm is able to effectively utilize the co-dependencies among the different shapes to aid in the segmentation process. It is able to capture a wide range of shape variability despite being a parametric shape-model. And finally, the algorithm is robust to large amounts of additive noise. We demonstrate the utility of this segmentation framework by applying it to a medical application: the segmentation of the prostate gland, the rectum, and the internal obturator muscles for MR-guided prostate brachytherapy.
We demonstrate that classical Business Process Reengineering (BPR) methods can be successfully applied to Computer Aided Surgery while increasing safety and efficiency of the overall procedure through an integrated Workflow Management System. Computer guided Prostate Brachytherapy, as a sophisticated treatment by an interdisciplinary team, is perfectly suited to apply our method. Detailed suggestions for improvement of the whole procedure could be derived by our modified BPR method.
BACKGROUND: One of the diagnostic imaging hallmarks of MS is the uptake of IV administered contrast material in new lesions in the brain, signaling blood-brain barrier breakdown and active inflammation. Many clinical drug trials are designed based on the assumption that lesion enhancement on MRI remains visible on average for 1 month. For practical reasons, few serial MRI studies of patients with MS have been performed at intervals shorter than 4 weeks.
METHODS: The authors performed a year-long longitudinal study in 26 patients with relapsing-remitting MS (RRMS), which comprised an initial phase of MRI follow-up at weekly intervals for 8 weeks, followed by imaging every other week for another 16 weeks, and monthly thereafter. They present a quantitative analysis (using a supervised interactive thresholding procedure) of new enhancing lesions appearing during the first 6 weeks in this cohort and evaluated from the time of first detection until enhancement was no longer seen.
RESULTS: The average duration of Gd-DTPA enhancement in individual new lesions was 3.07 weeks (median, 2 weeks). Significant correlations were demonstrated between the duration of contrast enhancement or initial growth rates and lesion volumes. Different lesions in the same patient appeared to develop largely independent of each other and demonstrated a large range in the duration of enhancement during the acute phase of their evolution.
CONCLUSIONS: The average duration of blood-brain barrier impairment in RRMS is shorter than earlier estimates. Early lesion growth parameters may predict final lesion size. Within-patient heterogeneity of lesion evolution suggests that individual lesions develop independently.
OBJECTIVE: Smaller temporal lobe cortical gray matter volumes, including the left superior temporal gyrus, have been reported in magnetic resonance imaging (MRI) studies of patients with chronic schizophrenia and, more recently, in patients with first-episode schizophrenia. However, it remains unknown whether there are progressive decreases in temporal lobe cortical gray matter volumes in patients with first-episode schizophrenia and whether similarly progressive volume decreases are present in patients with affective psychosis.
METHOD: High-spatial-resolution MRI scans at initial hospitalization and 1.5 years later were obtained from 13 patients with first-episode schizophrenia, 15 patients with first-episode affective psychosis (mainly manic), and 14 healthy comparison subjects. MRI volumes were calculated for gray matter of superior temporal gyrus and for the amygdala-hippocampal complex.
RESULTS: Patients with first-episode schizophrenia showed significant decreases in gray matter volume over time in the left superior temporal gyrus compared with patients with first-episode affective psychosis or healthy comparison subjects. This progressive decrease was more pronounced in the posterior portion of the left superior temporal gyrus (mean=9.6%) than in the anterior portions (mean=8.4%). No group differences in the rate of change over time were present in other regions.
CONCLUSIONS: These findings demonstrate a progressive volume reduction of the left posterior superior temporal gyrus gray matter in patients with first-episode schizophrenia but not in patients with first-episode affective psychosis.
Functional Magnetic Resonance Imaging (fMRI) studies are derived from a time series of Echo-Planar images (EPIs). Compared to conventional Magnetic Resonance Images (MRIs), EPIs are of relatively poor quality for discerning anatomic features and are often registered with corresponding MRIs to map brain activity to neuroanatomy. In this paper we demonstrate the utility of a technique to register an EPI-MRI pair by minimizing the discrepancy between its joint intensity probability mass function (PMF) and a previously learned one for a properly registered EPI-MRI pair, using the Kullback-Leibler Distance (KLD). In probing experiments Joint Entropy (JE) and Mutual Information showed significant bias relative to KLD along the axial direction and JE along a rotation axis. A comparison of searches using random starting poses showed KLD to have lower final pose errors than JE. Results of variation on parameters of the KLD based EPI-MRI registration technique are also presented.
Current clinical practice in the premature infant with posthaemorrhagic ventricular dilatation (PHVD) includes drainage of cerebrospinal fluid (CSF). This case study used advanced volumetric three dimensional magnetic resonance imaging to document the impact of CSF removal on the volume of regional brain tissues in a premature infant with PHVD. The removal of a large volume of CSF was associated with an identical reduction in CSF volume, but more dramatically with a significant increase in the regional volumes of cortical grey matter and myelinated white matter. The alterations in cerebral cortical grey matter and myelinated white matter volumes may provide insight into the established association of PHVD with deficits in cognitive and motor functions.