Quantification of image contrast of infarcts on computed tomography scans.

Introduction Accurate identification of infarcts in non-contrast computed tomography (NC-CT) scans of the brain is fundamental in the diagnosis and management of patients with stroke. Quantification of image contrast properties at the boundaries of ischemic infarct regions in NC-CT can contribute to a more precise manual or automatic delineation of these regions. Here we explore these properties quantitatively. Methods We retrospectively investigated 519 NC-CT studies of 425 patients with clinically confirmed ischemic strokes. The average and standard deviation (SD) of patients' age was 67.5 ± 12.4 years and the average(median)±SD time from symptoms onset to NC-CT examination was 27.4(12)±35.7 h. For every scan with an ischemic lesion identified by experts, the image contrast of the lesion vs. normal surrounding parenchyma was calculated as a difference of mean Hounsfield Unit (HU) of 1-5 consecutive voxels (the contrast window width) belonging to the lesion and to the parenchyma. This contrast was calculated at each single voxel of ischemic lesion boundaries (previously delineated by the experts) in horizontal and vertical directions in each image. The distributions of obtained horizontal, vertical and both contrasts combined were calculated among all 519 NC-CTs. Results The highest applicative contrast window width was identified as 5 voxels. The ischemic infarcts were found to be characterized by 6.60 HU, 8.28 HU and 7.55 HU mean values for distributions of horizontal, vertical and combined contrasts. Approximately 40-50% of the infarct boundary voxels were found to refer to the image contrast below 5 HU. Conclusion Low image contrast of ischemic lesions prevents accurate delineation of the infarcts in NC-CT.

Increased body mass index makes an impact on brain white-matter integrity in adults with remitted first-episode mania.

BACKGROUND: Obesity is increasingly prevalent in bipolar disorder (BD) but data about the impact of elevated body mass index (BMI) on brain white-matter integrity in BD are sparse. Based on extant literature largely from structural magnetic resonance imaging (MRI) studies, we hypothesize that increased BMI is associated with decreased fractional anisotropy (FA) in the frontal, temporal, parietal and occipital brain regions early in the course of BD. METHOD: A total of 26 euthymic adults (12 normal weight and 14 overweight/obese) with remitted first-episode mania (FEM) and 28 controls (13 normal weight and 15 overweight/obese) matched for age, handedness and years of education underwent structural MRI and diffusion tensor imaging scans. RESULTS: There are significant effects of diagnosis by BMI interactions observed especially in the right parietal lobe (adjusted F(1,48) = 5.02, p = 0.030), occipital lobe (adjusted F(1,48) = 10.30, p = 0.002) and temporal lobe (adjusted F(1,48) = 7.92, p = 0.007). Specifically, decreased FA is found in the right parietal (F(1,48) = 5.864, p = 0.023) and occipital lobes (F(1,48) = 4.397, p = 0.047) within overweight/obese patients compared with normal-weight patients with FEM. Compared with overweight/obese controls, decreased FA is observed in right parietal (F(1,48) = 6.708, p = 0.015), temporal (F(1,48) = 10.751, p = 0.003) and occipital (F(1,48) = 9.531, p = 0.005) regions in overweight/obese patients with FEM. CONCLUSIONS: Our findings suggest that increased BMI affects temporo-parietal-occipital brain white-matter integrity in FEM. This highlights the need to further elucidate the relationship between obesity and other neural substrates (including subcortical changes) in BD which may clarify brain circuits subserving the association between obesity and clinical outcomes in BD.

Stroke atlas: a 3D interactive tool correlating cerebrovascular pathology with underlying neuroanatomy and resulting neurological deficits.

Understanding stroke-related pathology with underlying neuroanatomy and resulting neurological deficits is critical in education and clinical practice. Moreover, communicating a stroke situation to a patient/family is difficult because of complicated neuroanatomy and pathology. For this purpose, we created a stroke atlas. The atlas correlates localized cerebrovascular pathology with both the resulting disorder and surrounding neuroanatomy. It also provides 3D display both of labeled pathology and freely composed neuroanatomy. Disorders are described in terms of resulting signs, symptoms and syndromes, and they have been compiled for ischemic stroke, hemorrhagic stroke, and cerebral aneurysms. Neuroanatomy, subdivided into 2,000 components including 1,300 vessels, contains cerebrum, cerebellum, brainstem, spinal cord, white matter, deep grey nuclei, arteries, veins, dural sinuses, cranial nerves and tracts. A computer application was developed comprising: 1) anatomy browser with the normal brain atlas (created earlier); 2) simulator of infarcts/hematomas/aneurysms/stenoses; 3) tools to label pathology; 4) cerebrovascular pathology database with lesions and disorders, and resulting signs, symptoms and/or syndromes. The pathology database is populated with 70 lesions compiled from textbooks. The initial view of each pathological site is preset in terms of lesion location, size, surrounding surface and sectional neuroanatomy, and lesion and neuroanatomy labeling. The atlas is useful for medical students, residents, nurses, general practitioners, and stroke clinicians, neuroradiologists and neurologists. It may serve as an aid in patient-doctor communication helping a stroke clinician explain the situation to a patient/family. It also enables a layman to become familiarized with normal brain anatomy and understand what happens in stroke.

Bridging neuroanatomy, neuroradiology and neurology: three-dimensional interactive atlas of neurological disorders.

Understanding brain pathology along with the underlying neuroanatomy and the resulting neurological deficits is of vital importance in medical education and clinical practice. To facilitate and expedite this understanding, we created a three-dimensional (3D) interactive atlas of neurological disorders providing the correspondence between a brain lesion and the resulting disorder(s). The atlas contains a 3D highly parcellated atlas of normal neuroanatomy along with a brain pathology database. Normal neuroanatomy is divided into about 2,300 components, including the cerebrum, cerebellum, brainstem, spinal cord, arteries, veins, dural sinuses, tracts, cranial nerves (CN), white matter, deep gray nuclei, ventricles, visual system, muscles, glands and cervical vertebrae (C1-C5). The brain pathology database contains 144 focal and distributed synthesized lesions (70 vascular, 36 CN-related, and 38 regional anatomy-related), each lesion labeled with the resulting disorder and associated signs, symptoms, and/or syndromes compiled from materials reported in the literature. The initial view of each lesion was preset in terms of its location and size, surrounding surface and sectional (magnetic resonance) neuroanatomy, and labeling of lesion and neuroanatomy. In addition, a glossary of neurological disorders was compiled and for each disorder materials from textbooks were included to provide neurological description. This atlas of neurological disorders is potentially useful to a wide variety of users ranging from medical students, residents and nurses to general practitioners, neuroanatomists, neuroradiologists and neurologists, as it contains both normal (surface and sectional) brain anatomy and pathology correlated with neurological disorders presented in a visual and interactive way.

Three-dimensional interactive atlas of cranial nerve-related disorders.

Anatomical knowledge of the cranial nerves (CN) is fundamental in education, research and clinical practice. Moreover, understanding CN-related pathology with underlying neuroanatomy and the resulting neurological deficits is of vital importance. To facilitate CN knowledge anatomy and pathology understanding, we created an atlas of CN-related disorders, which is a three-dimensional (3D) interactive tool correlating CN pathology with the underlying surface and sectional neuroanatomy as well as the resulting neurological deficits. A computer platform was developed with: 1) anatomy browser along with the normal brain atlas (built earlier); 2) simulator of CN lesions; 3) tools to label CN-related pathology; and 4) CN pathology database with lesions and disorders, and the resulting signs, symptoms and/or syndromes. The normal neuroanatomy comprises about 2,300 3D components subdivided into modules. Cranial nerves contain more than 600 components: all 12 pairs of cranial nerves (CN I - CN XII) and the brainstem CN nuclei. The CN pathology database was populated with 36 lesions compiled from clinical textbooks. The initial view of each disorder was preset in terms of lesion location and size, surrounding surface and sectional neuroanatomy, and disorder and neuroanatomy labeling. Moreover, path selection from a CN nucleus to a targeted organ further enhances pathology-anatomy relationships. This atlas of CN-related disorders is potentially useful to a wide variety of users ranging from medical students and residents to general practitioners, neuroradiologists and neurologists, as it contains both normal brain anatomy and CN-related pathology correlated with neurological disorders presented in a visual and interactive way.

Automatic segmentation of ventricular cerebrospinal fluid from ischemic stroke CT images.

Accurate segmentation of ventricular cerebrospinal fluid (CSF) regions in stroke CT images is important in assessing stroke patients. Manual segmentation is subjective, time consuming and error prone. There are currently no methods dedicated to extracting ventricular CSF regions in stroke CT images. 102 ischemic stroke CT scans (slice thickness between 3 and 6 mm, voxel size in the axial plane between 0.390 and 0.498 mm) were acquired. An automated template-based algorithm is proposed to extract ventricular CSF regions which accounts for the presence of ischemic infarct regions, image noise, and variations in orientation. First, template VT(2) is registered to the scan using landmark-based piecewise linear scaling and then template VT(1) is used to further refine the registration by partial segmentation of the fourth ventricle. A region of interest (ROI) is found using the registered VT(2). Automated thresholding is then applied to the ROI and the artifacts are removed in the final phase. Sensitivity, dice similarity coefficient, volume error, conformity and sensibility of segmentation results were 0.74 ± 0.12, 0.8 ± 0.09, 0.16 ± 0.11, 0.45 ± 0.39, 0.88 ± 0.09, respectively. The processing time for a 512 × 512 × 30 CT scan takes less than 30 s on a 2.49 GHz dual core processor PC with 4 GB RAM. Experiments with clinical stroke CT scans showed that the proposed algorithm can generate acceptable results in the presence of noise, size variations and orientation differences of ventricular systems and in the presence of ischemic infarcts.

A Simple and Fast Method of 3D Registration and Statistical Landmark Localization for Sparse Multi-Modal/Time-Series Neuroimages Based on Cortex Ellipse Fitting.

Existing methods of neuroimage registration typically require high quality scans and are time-consuming. We propose a simple and fast method which allows intra-patient multi-modal and time-series neuroimage registration as well as landmark identification (including commissures and superior/inferior brain landmarks) for sparse data. The method is based on elliptical approximation of the brain cortical surface in the vicinity of the midsagittal plane (MSP). Scan registration is performed by a 3D affine transformation based on parameters of the cortex elliptical fit and by aligning the MSPs. The landmarks are computed using a statistical localization method based on analysis of 53 structural scans without detectable pathology. The method is illustrated for multi-modal registration, analysis of hemorrhagic stroke time series, and ischemic stroke follow ups, as well as for localization of hardly visible or not discernible landmarks in sparse neuroimages. The method also enables a statistical localization of landmarks in sparse morphological/non-morphological images, where landmark points may be invisible.

A Brain Parenchyma Model-Based Segmentation of Intraventricular and Intracerebral Haemorrhage in CT Scans.

Accurate quantification of haemorrhage volume in a computed tomography (CT) scan is critical in the management and treatment planning of intraventricular (IVH) and intracerebral haemorrhage (ICH). Manual and semi-automatic methods are laborious and time-consuming limiting their applicability to small data sets. In clinical trials measurements are done at different locations and on a large number of data; an accurate, consistent and automatic method is preferred. A fast and efficient method based on texture energy for identification and segmentation of hemorrhagic regions in the CT scans is proposed. The data set for the study was obtained from CLEAR-IVH clinical trial phase III (41 patients' 201 sequential CT scans from ten different hospitals, slice thickness 2.5-10 mm and from different scanners). The DICOM data were windowed, skull stripped, convolved with textural energy masks and segmented using a hybrid method (a combination of thresholding and fuzzy c-means). Artifacts were removed by statistical analysis and morphological processing. Segmentation results were compared with the ground truth. Descriptive statistics, Dice statistical index (DSI), Bland-Altman and mean difference analysis were carried out. The median sensitivity, specificity and DSI for slice identification and haemorrhage segmentation were 86.25%, 100%, 0.9254 and 84.90%, 99.94%, 0.8710, respectively. The algorithm takes about one minute to process a scan in MATLAB(®). A hybrid method-based volumetry of haemorrhage in CT is reliable, observer independent, efficient, reduces the time and labour. It also generates quantitative data that is important for precise therapeutic decision-making.

The human brain in 1700 pieces: design and development of a three-dimensional, interactive and reference atlas.

As the human brain is the most complex living organ, constructing its detailed model with exploration capabilities in a form of an atlas is a challenge. Our overall goal is to construct an advanced, detailed, parcellated, labeled, accurate, interactive, three-dimensional (3D), and scalable whole human brain atlas of structure, vasculature, tracts and systems. The objectives of this work are three-fold; to present: (1) method of atlas design and development including design principles, accuracy requirements, atlas content, architecture, functionality, user interface, and customized tools; (2) creation of an atlas of structure and systems including its modeling method and validation; and (3) integration of this atlas with the cerebrovasculature and tracts created earlier. The atlas is created from multiple in vivo 3/7 T scans. Its design based on "pyramidal principle" enables scalability while preserving design principles and exploits interaction paradigm "from blocks to brain". The atlas contains (1) navigator with modules for system/object/object state management, interaction, user interfacing, and rendering; and (2) brain model with cerebrum, cerebellum, brainstem, spinal cord, white matter, deep structures, systems, ventricles, arteries, veins, sinuses, and tracts. The brain model is parcellated, labeled, consistent, realistic, of high resolution, polygonal/volumetric, dissectible, extendable, and deformable. It has over 1700 3D components. The atlas has sub-voxel accuracy of 0.1mm and the smallest vessels of 80 µm. Brain exploration includes dynamic scene composition, manipulation-independent 3D labeling, interaction combined with animation, meta-labeling, and quantification. This atlas is useful in education, research, and clinical applications. It can potentially be foundation for a multi-level molecular-cellular-anatomical-physiological-behavioral platform.

Three-dimensional reference and stereotactic atlas of human cerebrovasculature from 7Tesla.

Although there exist some reference and stereotactic anatomical human brain atlases, there is no equivalent cerebrovascular atlas. A 3D reference atlas of the human cerebrovasculature that is interactive, stereotactic, very detailed, completely parcellated, fully labeled, extendable, dissectible, deformable, and user friendly, is needed in education, training, research, and clinical applications. We constructed a sophisticated and very detailed cerebrovascular atlas with 1325 vascular segments, the smallest of 80 µm in diameter. The atlas was created from multiple 3 and 7T scans by employing tools and methods developed in our lab. In contrast to a sort of "sampled" vascular geometry cited in the literature, we provide information about the "continuous" geometry of the vascular model measurable in 3D at every location and along any vascular segment for both hemispheres. This cerebrovascular atlas was validated taking into account domain knowledge from various fields including angiography, neurosurgery, neuroradiology, (clinical) neuroanatomy, and terminology. Validation was considered as a confirmation of the created vascular model to a typical (conventional) cerebrovascular system with nearly average dimensions. It was done in terms of vessel subdivision, origin, course, surrounding anatomy, patterns, length, diameter, and branches. This validation demonstrates that the constructed cerebrovascular model generally represents a normal, typical cerebrovasculature with its dimensions close to average. This 7T atlas along with the vascular model is a rich source of knowledge about the human cerebrovasculature. The atlas is applicable in education, research, and clinical applications; it has already been applied in deep brain stimulation.

Low free triiodothyronine levels are related to poor prognosis in acute ischemic stroke.

BACKGROUND: Neuroendocrine changes are important processes which accompany critical illness, however, the number of clinical studies concentrating on the role of thyroid gland hormones in stroke pathogenesis is relatively small. The aim of this prospective study was to investigate the relation between free triiodothyronine (fT3) levels and the prognosis of patients with stroke. METHODS: The prospective study included 387 patients with acute (<24 h of symptoms onset) ischemic stroke consecutively admitted to Stroke Units. The subjects with known conditions that could interfere with thyroid gland metabolism were excluded. We analyzed: the routine blood tests, fT3, free thyroxine (fT4), thyroid-stimulating hormone (TSH) levels, unenhanced CT scans, initial clinical status (NIH Stroke Scale, NIHSS), 30- and 360- days outcome (modified Rankin Scale-mRS) and calculated the survival rate. RESULTS: A higher NIHSS score was in the 1 (st) fT3 levels tertile, whereas a lower in the 3 (rd) fT3 levels tertile (p=0.006). The 30- and 360-days mRS scores showed that patients in the lowest fT3 tertile had more severe neurological impairment than those in the highest tertile (p=0.001 and p=0.03, respectively). A 1-year mortality of the patients with the first tertile fT3 levels was significantly higher than that of the patients with the third tertile hormone levels (p=0.008). Additionally, subjects with fT3 level in the lowest tertile demonstrated higher WBC counts and the ventricular system on Computed Tomography of head performed on admission to hospital was statistically more frequent compressed than that in the patients with fT3 level in the highest tertile (p=0.02 and p=0.03, respectively). CONCLUSION: In acute stroke patients lower free T3 levels are an important factor related to unfavorable outcome, i. e., severe disability and death.

Quantitative analysis of human masticatory muscles using magnetic resonance imaging.

OBJECTIVES: The objective of this study was to quantitatively evaluate the correlation between left and right masticatory muscle volumes in normal subjects. METHODS: Contiguous 1 mm MR scans were obtained of 12 normal adult subjects aged 20-25 years using a Siemens 1.5 T MR scanner. The volumes of the human masticatory muscles (masseter, lateral and medial pterygoid) were measured from the scans using our previously proposed method. To test for inter- and intraobserver reproducibility, measurements were performed by two users on two separate occasions, with a span of 2 weeks between them and with the previous results blinded. Good inter- and intraobserver reproducibility was achieved in our study. RESULTS: The mean volumes for left and right masseters, and lateral and medial pterygoids were 29.54 cm3, 29.65 cm3, 9.47 cm3, 10.23 cm3, 8.69 cm3 and 8.92 cm3, respectively. The Pearson correlation coefficients between the volumes of the left and right masseters, lateral and medial pterygoids are 0.969, 0.906 and 0.924, respectively. CONCLUSIONS: The computed volumes of the masticatory muscles show a strong correlation between the volumes of the left and right masseters, and lateral and medial pterygoids for normal adult subjects. The total masticatory muscle volume on the left and right sides of normal subjects is similar.

3D segmentation and quantification of a masticatory muscle from MR data using patient-specific models and matching distributions.

A method is proposed for 3D segmentation and quantification of the masseter muscle from magnetic resonance (MR) images, which is often performed in pre-surgical planning and diagnosis. Because of a lack of suitable automatic techniques, a common practice is for clinicians to manually trace out all relevant regions from the image slices which is extremely time-consuming. The proposed method allows significant time savings. In the proposed method, a patient-specific masseter model is built from a test dataset after determining the dominant slices that represent the salient features of the 3D muscle shape from training datasets. Segmentation is carried out only on these slices in the test dataset, with shape-based interpolation then applied to build the patient-specific model, which serves as a coarse segmentation of the masseter. This is first refined by matching the intensity distribution within the masseter volume against the distribution estimated from the segmentations in the dominant slices, and further refined through boundary analysis where the homogeneity of the intensities of the boundary pixels is analyzed and outliers removed. It was observed that the left and right masseter muscles' volumes in young adults (28.54 and 27.72 cm(3)) are higher than those of older (ethnic group removed) adults (23.16 and 22.13 cm(3)). Evaluation indicates good agreement between the segmentations and manual tracings, with average overlap indexes for the left and right masseters at 86.6% and 87.5% respectively.

Quantification and visualization of three-dimensional inconsistency of the ventrointermediate nucleus of the thalamus in the Schaltenbrand-Wahren brain atlas.

BACKGROUND: This work quantifies and visualises 3D inconsistencies of the ventrointermediate nucleus (VIM) of the thalamus, including the VIM externum (VIMe) and VIM internum (VIMi), in the Schaltenbrand-Wahren (SW) brain atlas. METHOD: For each VIM, VIMe, VIMi the 3D models, 3D-A, 3D-C and 3D-S were reconstructed from the SW axial, coronal and sagittal microseries, respectively, by applying a shape-based method. All 3D models, placed in the SW coordinate system, were compared quantitatively in terms of location (centroids), size (volumes), shape (normalised eigen values), orientation (eigen vectors), and mutual spatial relationships (overlaps and inclusions). FINDINGS: The reconstructed 3D models differ significantly in location, size, shape, and inclusion rate. The centroid of 3D-A/VIM differs considerably from those of 3D-C/VIM and 3D-S/VIM. The difference between the centroids of 3D-C/VIM and 3D-S/VIM is in laterality only: that of 3D-C/VIM is located more medially (11.85 mm) than that of 3D-S/VIM (14.62 mm). 3D-A/VIM has the smallest volume (69.00 mm(3)); 3D-C/VIM is 3.71 and 3D-S/VIM 3.89 times larger. The overlap is also highly variable: 104.88 mm(3) for 3D-C/VIM with 3D-S/VIM, and very low (3.22 and 7.45 mm(3)) when 3D-A/VIM is involved. The highest inclusion rate is for 3D-C/VIM with 3D-S/VIM (39.10 and 40.97%) and the lowest for 3D-A/VIM with 3D-C/VIM (1.26 and 4.66%). The centroid of 3D-A/VIMe differs noticeably from those of 3D-C/VIMe and 3D-S/VIMe. The difference between the centroids of 3D-C/VIMe and 3D-S/VIMe is mainly in laterality: that of 3D-C/VIMe is located more medially (12.91 mm) than that of 3D-S/VIMe (16.65 mm). 3D-A/VIMe has the smallest volume (49.87 mm(3)); 3D-S/VIMe is 3.24 and 3D-C/VIMe 3.36 times larger. The overlap sizes are low: 32.72 mm(3) for 3D-C/VIMe with 3D-S/VIMe, and very low (1.32 and 2.01 mm(3)) when 3D-A/VIMe is involved. The inclusion rates are also low: the highest is for 3D-C/VIMe with 3D-S/VIMe (19.53 and 20.29%) and the lowest for 3D-A/VIMe with 3D-C/VIMe (1.19 and 4.01%). Lateral scaling of the coronal microseries by 1.2897 to match the 3D-C/VIMe and 3D-S/VIMe centroids increases the inclusion rates for the sagittal microseries by more than twice. The volume of scaled 3D-C enlarges to 216.24 mm(3) which is 1.34 bigger than that of 3D-S. There are substantial differences among the centroids of 3D-A/VIMi, 3D-C/VIMi and 3D-S/VIMi. The centroid of 3D-A/VIMi is located more anteriorly (-1.92 mm) than that of 3D-C/VIMi (-5.02 mm). The centroid of 3D-A/VIMi is located more ventrally (2.88 mm) than those of 3D-C/VIMi and 3D-S/VIMi (each at 5.34 mm). 3D-A/VIMi has the smallest volume (19.75 mm(3)); 3D-S/VIMi is 3.23 and 3D-C/VIMi 4.30 times larger. 3D-A/VIMi practically does not overlap with 3D-C/VIMi and 3D-S/VIMi. The inclusion rates for 3D-C/VIMi with 3D-S/VIMi are medium (32.63 and 43.43%). CONCLUSION: Each VIM, VIMe, VIMi as reconstructed from the SW atlas has a significant 3D inaccuracy within each orientation and across them. Therefore, absolute and direct reliance on the original SW atlas is unreliable and unsafe, and this atlas has to be used with great care and understanding of its strengths and limitations.

Estimation and presentation of blood flow and velocity from angiographic scans in the human cerebral arterial system.

This paper provides an overview of blood flow in the arterial system and aims to estimate the blood velocity from cerebral angiography scans without having acquired data on velocity by using Murray's Law. The estimation technique post-processes the scan and provides crucial 3D visual data for the development of a visualization program of the blood flow in the human brain.

Medical image segmentation using watershed segmentation with texture-based region merging.

The use of the watershed algorithm for image segmentation is widespread because it is able to produce a complete division of the image. However, it is susceptible to over-segmentation and in medical image segmentation, this meant that that we do not have good representations of the anatomy. We address this issue by thresholding the gradient magnitude image and performing post-segmentation merging on the initial segmentation map. The automated thresholding technique is based on the histogram of the gradient magnitude map while the post-segmentation merging is based on the similarity in textural features (namely angular second moment, contrast, entropy and inverse difference moment) belonging to two neighboring partitions. When applied to the segmentation of various facial anatomical structures from magnetic resonance (MR) images, the proposed method achieved an overlap index of 92.6% compared to manual contour tracings. It is able to merge more than 80% of the initial partitions, which indicates that a large amount of over-segmentation has been reduced. Results produced using watershed algorithm with and without the proposed and proposed post-segmentation merging are presented for comparisons.

Masseter segmentation using an improved watershed algorithm with unsupervised classification.

The watershed algorithm always produces a complete division of the image. However, it is susceptible to over-segmentation and sensitivity to false edges. In medical images this leads to unfavorable representations of the anatomy. We address these drawbacks by introducing automated thresholding and post-segmentation merging. The automated thresholding step is based on the histogram of the gradient magnitude map while post-segmentation merging is based on a criterion which measures the similarity in intensity values between two neighboring partitions. Our improved watershed algorithm is able to merge more than 90% of the initial partitions, which indicates that a large amount of over-segmentation has been reduced. To further improve the segmentation results, we make use of K-means clustering to provide an initial coarse segmentation of the highly textured image before the improved watershed algorithm is applied to it. When applied to the segmentation of the masseter from 60 magnetic resonance images of 10 subjects, the proposed algorithm achieved an overlap index (kappa) of 90.6%, and was able to merge 98% of the initial partitions on average. The segmentation results are comparable to those obtained using the gradient vector flow snake.

Simplifying the exploration of volumetric images: development of a 3D user interface for the radiologist's workplace.

Volumetric imaging (computed tomography and magnetic resonance imaging) provides increased diagnostic detail but is associated with the problem of navigation through large amounts of data. In an attempt to overcome this problem, a novel 3D navigation tool has been designed and developed that is based on an alternative input device. A 3D mouse allows for simultaneous definition of position and orientation of orthogonal or oblique multiplanar reformatted images or slabs, which are presented within a virtual 3D scene together with the volume-rendered data set and additionally as 2D images. Slabs are visualized with maximum intensity projection, average intensity projection, or standard volume rendering technique. A prototype has been implemented based on PC technology that has been tested by several radiologists. It has shown to be easily understandable and usable after a very short learning phase. Our solution may help to fully exploit the diagnostic potential of volumetric imaging by allowing for a more efficient reading process compared to currently deployed solutions based on conventional mouse and keyboard.

A study on shape determinative slices for the masseter muscle.

In this paper, we localize the shape-determinative slices of the masseter, which plays a critical role in the mastication system, from magnetic resonance (MR) data sets for clinical purposes. Shape-based criteria were used to locate the candidates for determinative slices from training data. The localization process involves tracking of the centroid and detecting the locations where the structure of the masseter undergoes an abrupt change in orientation. Having determined all the candidates which satisfy the criteria, fuzzy-c-means (FCM) clustering technique was used to establish the determinative slices. Localization of these slices will facilitate the building of more accurate models. It will also allow for more accurate computerized extraction of the masseter from MR data. In our work here, a hybrid method to shape-based interpolation is used to build the masseter model from magnetic resonance (MR) data sets, and the mean overlap index (¿) achieved is 87.7%. Extraction of the masseter was carried out using our earlier proposed method and the mean ¿ achieved is 8.9%. This indicates good agreement between the results obtained using computerized technique and those contained using manual contour tracing.

Medical image segmentation using feature-based GVF snake.

We propose a feature-based GVF snake for medical image segmentation here. Feature-based criteria are introduced for the GVF snake to stop its iterations. Without these criteria, the GVF snake might continue its iterations even though it has converged at the targeted object and result in longer computational time. The feature here is the area of the targeted object. Our proposed method comprises of two stages, namely the training stage and the segmentation stage. In the training stage, we acquire prior knowledge on the relative area of the targeted object from training data. In the segmentation stage, the proposed feature-based GVF snake is applied to segment the object from the image after computing the estimated area of the targeted object. In our proposed method, the GVF snake stops its iterations when the area bounded by its propagation is approximately equal to the estimated area and when it undergoes little change over two consecutive iterations. To illustrate the effectiveness of our proposed method, we applied it to the segmentation of the masseter muscle, which is the strongest jaw muscle, from 2-D magnetic resonance (MR) images. Numerical evaluation done indicates good agreement between the computerized and manual segmentations, with mean overlap of 92%.