Electron Beam Tomography offers a fast method to acquire large numbers of X-ray CT slices. Scanning is performed by sweeping an electron beam over a curved tungsten ring that encompasses the body and serves as a high speed X-ray source. This technology is now in use in more than 100 hospital and clinics and has enabled a number of important new diagnostic applications, especially for 3D imaging of organs and structures that are moving. This paper will describe recent results in chest imaging applications, coronary artery imaging, and 4D-imaging of cardiopulmonary structures. The technology is under continuous development and will lead to future machines with higher 3D and 4D resolution and speed.

Application of the Adaptive Multiple Feature Method (AMFM) to identify early changes in a smoking population is discussed. This method was specifically applied to determine if differences in CT images of smokers (with normal lung function) and non-smokers (with normal lung function) could be found through computerized texture analysis. Results demonstrated that these groups could be differentiated with over 80.0% accuracy. Further, differences on CT images between normal appearing lung from non-smokers (with normal lung function) and normal appearing lung from smokers (with abnormal lung function) were also investigated. These groups were differentiated with over 89.5% accuracy. In analyzing the whole lung region by region, the AMFM characterized 38.6% of a smoker lung (with normal lung function) as mild emphysema. We can conclude that the AMFM detects parenchymal patterns in the lungs of smokers which are different from normal patterns occurring in healthy non-smokers. These patterns could perhaps indicate early smoking-related changes.

The purpose of this work was to develop an automated technique for calculating dynamic lung attenuation changes, through a forced expiratory maneuver, as a measure of split lung function. A total of ten patients post single lung transplantation (SLT) for emphysema were imaged using an Electron Beam CT Scanner; three were studied twice following stent placement. A single-slice flow study, using 100 msec exposures and 3 mm collimation, was performed at the level of the anastomosis during a forced expiration. Images were acquired every 500 msec for the first 3 seconds and every second for the last 4 seconds. An automated, knowledge-based system was developed to segment the chest wall, mediastinum, large airways and lung parenchyma in each image. Knowledge of the expected size, shape, topology and X-ray attenuation of anatomical structures were used to guide image segmentation involving attenuation thresholding, region-growing and morphology. From the segmented left and right parenchyma, the system calculated median attenuation (HU) and cross-sectional areas. These results were plotted against time for both the native and transplanted lungs. In five patients, significant shift of the attenuation/time curve to the right (slower flow) was detected, although the end expiration attenuation was not different. Following stent placement the curve shifted back to the left (faster flow).

Microvascular red blood cell mean transit time is a crucial parameter underlying basic pulmonary physiology. Dynamic x-ray CT imaging during bolus radiopaque tracer injection offers the ability to make functional measurements throughout the lungs, but is not able to resolve individual microvascular beds. We have implemented a model-free Fast Fourier Transform deconvolution algorithm to extract the microvascular transport characteristics from the acquired time-intensity data. The deconvolved feeding arterial bolus input curves and corresponding regional pulmonary parenchymal 'response' functions provide measures of regional pulmonary tracer residence times, allowing calculation of microvascular transit times for different spatial regions of the pulmonary system. The acquired feeding (main) pulmonary artery and regional pulmonary parenchyma time-intensity curves were fit to gamma variate functions which were then sampled with a temporal resolution of 0.1 seconds. Deconvolution of the feeding arterial and regional parenchymal curves consistently results in bimodal regional residue functions. The two modes consist of a relatively large, sharp, narrow peak approximating a delta function followed by a smaller more dispersed curve. The sharp, narrow peak appears to be due to small artery inclusion in the sampled parenchymal region (partial volume effects). The magnitude of the dominant arterial peak decreases as sampling locations are moved from the less expanded dependent to the more expanded non-dependent lung regions of supine dogs. Mathematical separation of the two modes allowed isolation of the arterial and microvascular components. The shape and transit times of the putative microvascular components agree well with results from similar measurements via microfocal angiography and in vivo microscopy. Reconvolving the microvascular component with the input curve results in a corrected parenchymal curve representing the regional microvascular transport characteristics, free of arterial flow signal contamination. The corrected residue curves can then be used for non-invasive in vivo quantitation of regional organ microvascular transit times, volumes and flows in relation to the existing in vivo anatomy.

X-ray microfocal angiography provides a means of assessing regional microvascular perfusion parameters using residue detection of vascular indicators. As an application of this methodology, we studied the effects of alveolar hypoxia, a pulmonary vasoconstrictor, on the pulmonary microcirculation to determine changes in regional blood mean transit time, volume and flow between control and hypoxic conditions. Video x-ray images of a dog lung were acquired as a bolus of radiopaque contrast medium passed through the lobar vasculature. X-ray time-absorbance curves were acquired from arterial and microvascular regions-of-interest during both control and hypoxic alveolar gas conditions. A mathematical model based on indicator-dilution theory applied to image residue curves was applied to the data to determine changes in microvascular perfusion parameters. Sensitivity of the model parameters to the model assumptions was analyzed. Generally, the model parameter describing regional microvascular volume, corresponding to area under the microvascular absorbance curve, was the most robust. The results of the model analysis applied to the experimental data suggest a significant decrease in microvascular volume with hypoxia. However, additional model assumptions concerning the flow kinematics within the capillary bed may be required for assessing changes in regional microvascular flow and mean transit time from image residue data.

In many clinical applications, it is desirable to perform computed tomography (CT) scans during the peak of the contrast uptake of an organ. Typically, a patient is first scanned continuously (with a low technique) at a fixed location and a series of CT images is generated. The average CT number within a predefined region of interest (ROI) is calculated and plotted for all the acquired images. When the contrast uptake within the ROI is judged to be adequate by an operator, a scan command is initiated and normal CT scans are performed. The methodology has been shown to be quite effective in improving the image quality of the CT examinations. When the scheme is applied to body scans (chest or abdomen), however, two factors might affect the effectiveness of this approach. The first relates to the fact that patient motion is not always avoidable in body scans. As a result, the predefined ROI might not always register at the same anatomical location. Another factor is related to the fact that a significant delay is encountered between the initiation of a scan command and the actual helical scan, due to mechanical delay of the CT system, time to move the patient table to the desired location, and the patient prep-time. Because of the contrast agent can pass through a patient organ in a relatively short period of time (10 - 30 seconds), the inherit delay often leads to a suboptimal scan since the peak of the contrast uptake is missed. In order to overcome these difficulties, we propose two schemes. The location of ROI is adapted from one image to the next to ensure the best feature match. This is accomplished by performing correlation calculation between the selected ROI in the pre-injection image and various ROIs in the neighboring locations in the new image. The final ROI location is determined based on the peak of the correlation coefficient. We also use a predictive algorithm to estimate the future uptake of the contrast agent. At the start of the scan, the average CT numbers are measured and sent to the predictor. The predictor uses all the measured samples taken up to that time to predict the average CT number a few seconds, t, later. When the next sample becomes available, the predictor will compare its prediction with the actual measurement and modify its prediction parameters to best predict the future samples. Computer simulations and clinical experiments have shown that significant improvement in the optimization of the contrast uptake can be achieved with this method. As a result, the entire CT examination is better centered on the peak of the contrast curve and an optimal contrast enhancement in CT image scan be achieved.

Although from a respiratory point of view, compartmental volume change or lack of it is the most crucial variable, it has not been possible to measure the volume of chest wall compartments directly. Recently we developed a new method based on a optoelectronic motion analyzer that can give the three-dimensional location of many markers with the temporal and spatial accuracy required for respiratory measurements. Marker's configuration has been designed specifically to measure the volume of three chest wall compartments, the pulmonary and abdominal rib cage compartments and the abdomen, directly. However, it can not track the exact border between the two rib cage compartments (pulmonary and abdominal) which is determined by the cephalic extremity of the area of apposition of the diaphragm to the inner surface of the rib cage, and which can change systematically as a result of disease processes. The diaphragm displacement can be detected by ultrasonography. In the present study, we propose an integrated system able to investigate the relationships between external (chest wall) and internal (diaphragm) movements of the different respiratory structures by simultaneous external imaging with the optoelectronic system combined with internal kinematic imaging using ultrasounds. 2D digitized points belonging to the lower lung margin, taken from ultrasonographic views, are mapped into the 3D space, where chest wall markers are acquired. Results are shown in terms of accuracy of 3D probe location, relative movement between the probe and the body landmarks, dynamic relationships between chest wall volume and position of the diaphragm during quiet breathing, slow inspirations, relaxations and exercise.

The aim of the study 'Dynamic Chest Image Analysis' is to develop computer analysis and visualization methods for showing focal and general abnormalities of lung ventilation and perfusion based on a sequence of digital chest fluoroscopy frames collected at different phases of the respiratory/cardiac cycles in a short period of time. We have proposed a framework for ventilation study with an explicit ventilation model based on pyramid images. In this paper, we extend the framework to pulmonary perfusion study. A perfusion model and the truncated pyramid are introduced. The perfusion model aims at extracting accurate, geographic perfusion parameters, and the truncated pyramid helps in understanding perfusion at multiple resolutions and speeding up the convergence process in optimization. Three cases are included to illustrate the experimental results.

Virtual bronchoscopy is emerging as a means for assessing high-resolution 3D CT images of the chest. The central axes, or paths, of the airways can provide virtual-bronchoscopic systems with a logical reference frame for quantitation and navigation. Unfortunately, the manual and automatic methods proposed to date for determining these axes are either time- consuming, error prone, or provide imprecise results. We give a preliminary presentation of an adaptive automated approach for finding smooth central axes through the major airways. Using this method, we are able to extract multiple axes through a 3D image in only a few minutes for a typical 512 by 512 by 25 CT image. The method works on anisotropically sampled gray-scale images and requires no prior segmentation. We describe the method and present initial validation results for phantom, animal, and human images. Visual results are also provided using a virtual bronchoscopic system.

The intravascular use of CO₂ as a contrast medium for diagnosis of vaso-occlusive disease has the potential of improved safety and reduced cost compared to conventional contrast solutions. Optimal imaging using any gaseous contrast medium is strongly dependent on the fluid mechanics of pulsatile gas-liquid flow. Gas-liquid flow regimes can be designated by the visually identifiable patterns of annular flow, stratified flow, intermittent flow, and dispersed bubble flow. This paper presents an analytically developed flow regime map for gas-liquid pulsatile flow and preliminary experimental results testing the flow regime map predictions. Preliminary experimental results confirm analytical predictions: (1) Use of CO₂ for intervascular vaso- occlusive imaging under typical conditions may result in a periodic intermittent flow; (2) Periodic intermittent flow is encountered at lower liquid and gas flow rates than is steady- state intermittent flow; (3) Reliable prediction of the flow regime in coordination with precise CO2 delivery can be used to prescribe a flow regime optimal for intravascular vaso- occlusive imaging.

To realize a non-contact, non-invasive and fast measurement of skin blood flow, we have developed the laser speckle contrast analysis (LASCA) technique. The LASCA method is a spatial domain method, based on the aggregate of pixels composing a captured laser speckle image. The contrast calculation operates directly on these pixels. In this paper, we present the computer algorithms to achieve a real-time solution for monitoring capillary blood flow and velocity. First, we present an improved naive LASCA algorithm with the running time O(k²n), where n is the total number of pixels and k X k represents the size of the subimage considered. Then, we describe a fast LASCA algorithm, which takes time O(kn) to calculate the local contrasts. Finally, we use the fast sequential algorithm to design the first LASCA parallel algorithm to run on the CREW PRAM with the running time O(k/p n), where p is the number of processors. Experimental result shows that, to process a laser speckle image with the size of 640 X 480, it takes only about one second using the fast LASCA algorithm. Furthermore, our parallel algorithm is easily implemented to run under the Windows NT environment by using multi-threads technique.

Atherosclerosis begins in childhood with the accumulation of lipid in the intima of arteries to form fatty streaks, advances through adult life when occlusive vascular disease may result in coronary heart disease, stroke and peripheral vascular disease. Non-invasive B-mode ultrasound has been found useful in studying risk factors in the symptom-free population. Large amount of data is acquired from continuous imaging of the vessels in a large study population. A high quality brachial vessel diameter measurement method is necessary such that accurate diameters can be measured consistently in all frames in a sequence, across different observers. Though human expert has the advantage over automated computer methods in recognizing noise during diameter measurement, manual measurement suffers from inter- and intra-observer variability. It is also time-consuming. An automated measurement method is presented in this paper which utilizes quality assurance approaches to adapt to specific image features, to recognize and minimize the noise effect. Experimental results showed the method's potential for clinical usage in the epidemiological studies.

A novel, fully automatic, multi-step segmentation method has been developed to distinguish different tissues in human limbs. As input data conventional transverse T1-weighted MR scans were used providing a good contrast between tissues of interest and background. Hence, a scalar (monospectral) histogram approach employing statistical distribution models of gray values has been chosen. The identification of the tissues becomes possible by alternating different levels of histogram-based thresholding techniques and sequences of morphologic operations in conjunction with a-priori anatomic knowledge. By using morphologic operations and generic definitions of the anatomic structures found in the limbs, specific parts of the histogram can be identified which otherwise impede the automatic histogram interpretation. To correct partial volume effects, which may corrupt the different modes of the histogram, gradient thresholding was established. The proposed method was validated on both phantoms and ten healthy volunteers. Using this method, a study was performed to investigate whether there is a contribution in reactive hyperemic flow induced by differences in forearm composition.

Previously developed empirical analyses of dynamic contrast- enhanced MR imaging (DEMRI) studies have provided a more quantitative and accurate measure of solid tumor response. However, empirically based methods do not generalize easily to other solid tumors, and changes in the parameters during therapy are hard to relate to physiological mechanisms. This study compares the kinetic parameters from a two-compartment pharmacokinetic (PK) model of MR contrast agent accumulation with disease free survival rates after surgery and investigates the serial changes in these parameters over the course of therapy in thirty-five patients with osteosarcoma. The PK model allowed us to directly determine the relationship between the first-order permeability rate constant (k₂₁) of the model and histologic assessment of response. A Cox proportional hazards model was used to compare the average k₂₁ immediately prior to surgery and was determined to be significantly related to disease free survival. A linear regression analysis between the average k₂₁ at presentation and the resulting change in average k₂₁ during therapy revealed a statistically significant relationship corresponding to greater delivery of contrast agent. Larger regional access at presentation corresponded to larger decreases in access during therapy, which is consistent with the hypothesis that greater regional access at presentation should correspond to greater response to the chemotherapeutic agents.

A new technique to measure local planar strain in left ventricular myocardium using two-dimensional tagged MR images is presented. This new technique is computationally fast, is fully automated, and generates dense motion estimates. It is based on using a 1-1 SPAMM tag pattern which comprise several one-dimensional sinusoidal tag patterns at different frequencies. A local deformation of the myocardium produces a variation in the local frequencies of these patterns, which can be used to compute strain components in the image plane. Local frequency is measured by scanning certain spectral peaks to create complex images, for which the local frequency is the gradient of the angle associated with their complex data points. The method is demonstrated using both simulations and real tagged MR images, and a discussion of these results and of directions for future research is provided.

This manuscript documents the alteration of the heart model of the MCAT phantom to better represent cardiac motion. The objective of the inclusion of motion was to develop a digital simulation of the heart such that the impact of cardiac motion on single photon emission computed tomography (SPECT) imaging could be assessed and methods of quantitating cardiac function could be investigated. The motion of the dynamic MCAT's heart is modeled by a 128 time frame volume curve. Eight time frames are averaged together to obtain a gated perfusion acquisition of 16 time frames and ensure motion within every time frame. The position of the MCAT heart was changed during contraction to rotate back and forth around the long axis through the center of the left ventricle (LV) using the end systolic time frame as turning point. Simple respiratory motion was also introduced by changing the orientation of the heart model in a 2 dimensional (2D) plane with every time frame. The averaging effect of respiratory motion in a specific time frame was modeled by randomly selecting multiple heart locations between two extreme orientations. Non-gated perfusion phantoms were also generated by averaging over all time frames. Maximal chamber volumes were selected to fit a profile of a normal healthy person. These volumes were changed during contraction of the ventricles such that the increase in volume in the atria compensated for the decrease in volume in the ventricles. The myocardium were modeled to represent shortening of muscle fibers during contraction with the base of the ventricles moving towards a static apex. The apical region was modeled with moderate wall thinning present while myocardial mass was conserved. To test the applicability of the dynamic heart model, myocardial wall thickening was measured using maximum counts and full width half maximum measurements, and compared with published trends. An analytical 3D projector, with attenuation and detector response included, was used to generate radionuclide projection data sets. After reconstruction a linear relationship was obtained between maximum myocardial counts and myocardium thickness, similar to published results. A numeric difference in values from different locations exist due to different amounts of attenuation present. Similar results were obtained for FWHM measurements. Also, a hot apical region on the polar maps without attenuation compensation turns into an apical defect with attenuation compensation. The apical decrease was more prominent in ED than ES due to the change in the partial volume effect. Both of these agree with clinical trends. It is concluded that the dynamic MCAT (dMCAT) phantom can be used to study the influence of various physical parameters on radionuclide perfusion imaging.

There is an established link between Left Ventricular (LV) geometry and its performance. As a consequence of ischemic heart disease and the attempt to relieve myocardial tissue stress, ventricle shape begins to distort from a conical to spherical geometry with a reduction in pumping efficiency of the chamber. If untreated, premature heart failure will result. To increase the changes of successful treatment it is obviously important for the benefit of the patient to detect these abnormalities as soon as possible. It is the development of a technique to characterize and quantify the shape of the left ventricle that is described here. The system described in this paper uses a novel helix model which combines the advantages of current two dimensional (2D) quantitative measures which provide limited information, with 3D qualitative methods which provide accurate reconstructions of the LV using computationally expensive rendering schemes. A phantom object and dog ventricle (normal/abnormal) were imaged and helical models constructed. The result are encouraging with differences between normal and abnormal ventricles in both diastole and systole able to be determined. Further work entails building a library of subjects in order to determine the relationship between ventricle geometry and quantitative measurements.

Velocity Magnetic Resonance (MR) images are a novel form of medical images. A special gradient-modulation technique is utilized to capture motion velocity of tissue and blood. As well as the tissue density image, there are also other images that depict the velocity components along axes defined relative to the plane of imaging. The images are of the cardiac region and are aligned with the short-axis of the left ventricle. We present the results of clustering cardiac image sequences using the Fuzzy c-Means (FCM) algorithm. Our paper demonstrates how the application of clustering to one frame in the cine sequence of images can be utilized in order to track reasonably well the contraction and relaxation of the Left Ventricle. Our paper shows that this imaging technique is generally accurate and certainly adds to the information already contained in the tissue density images.

This paper presents an integrated scheme to extract and reconstruct left ventricle chambers from CT volumetric image sequences. An accurate extraction of left ventricle chambers is a crucial step towards cardiac dynamics analysis based on image sequences, a very much desired non-invasive technique for heart disease diagnosis and monitoring. The integrated approach aims at solving two major problems in cardiac image segmentation: imaging related ambiguity and anatomy related ambiguity. The K-means clustering with Gibb's random field constraints is able to resolve the imaging related ambiguity to obtain robust segmentation even when the intensity of the left ventricle exhibits spatially varying distribution. The active contour models incorporating a priori shape knowledge is able to resolve the anatomy related ambiguity to estimate the valve that separates the left ventricle from left atrium and aorta but is indistinguishable in the given images due to motion and partial volume effects. The fusion of the clustering and active contour models enables an integrated reconstruction of left ventricle chambers from the CT image sequences. Preliminary results show that the proposed scheme can produce extracted left ventricle chambers that compare favorably with the manually delineated chambers by a skilled operator. However, this proposed scheme is fast and reproducible.

In this study, we explore the use of non-linear regression for model fitting of PET measured kinetics on a pixel-by-pixel basis for generating parametric images of micro-parameters of kinetic models. We evaluate quantitatively the noise propagation of two regression methods using computer simulated data, and examine the feasibility of generating parametric images for two different real PET studies -- a human FDG study and a monkey FDOPA study. The results demonstrated that general image-wise model fitting is practically feasible for dynamic PET studies.

Proton magnetic resonance spectroscopic imaging (1H MRSI) with volume pre-selection (i.e. by PRESS) or multislice 1H MRSI was used to investigate changes in brain metabolites in Alzheimer's disease, epilepsy, and amyotrophic lateral sclerosis. Examples of results from several ongoing clinical studies are provided. Multislice 1H MRSI of the human brain, without volume pre-selection offers considerable advantages over previously available techniques. Furthermore, MRI tissue segmentation and completely automated spectra curve fitting greatly facilitate quantitative data analysis. Future efforts will be devoted to obtaining full brain coverage and data acquisition at short spin echo times (TE less than 30 ms) for the detection of metabolites with short T2 relaxation times.

Imaging data, provided by functional magnetic resonance imaging (fMRI) and positron emission tomography (PET) has allowed insights into spatial organization of human cortical brain function. In the auditory system, a spatially-organized 'tonotopic' representation of frequency has been proposed. To provide an additional dimension to the description of functional organization, this study investigates the temporal signature of neuronal responses to auditory stimuli, detected using magnetoencephalography (MEG), to identify and characterize temporal encoding of stimulus attributes such as pitch and timbre. Stimuli elicited neuromagnetic fields with distinct peaks approximately 100 ms post-stimulus (M100). For sinusoidal tones frequency-dependent variation of the M100 latency was observed. This was mimicked using triangle and square waveforms, although the magnitude of the latency shift was attenuated. Amplitude-modulated tones and speech sounds demonstrated M100 latencies dominated by the carrier frequency and formats, and secondarily influenced by modulation frequency and fundamental, respectively. Information is encoded in the time domain of neuronal responses to auditory stimulation. Both the fundamental frequency and the timbre of a stimulus influence the specific latency of neuronal coherence. High temporal resolution recording of electromagnetic activity induced by stimuli allows new insights into the brain's functional encoding of presented information.

Surgical treatment of patients suffering from complex partial seizures requires the localization of the epileptogenic zone for surgical resection. Currently, clinicians utilize electroencephalography (EEG), psychological tests, and various neuroimaging modalities together to determine the location of this zone. We investigate the use of positron emission tomography (PET), magnetic resonance imaging (MRI), and magnetic resonance spectroscopy (MRS) in the presurgical workup and analysis of patients with complex partial seizures. The results of imaging studies of 25 patients are compared for lateralization accuracy and relative concordance.

Detailed cerebrovascular blood flow can be more accurately determined radiographically from the new droplet tracking method previously introduced by the authors than from standard soluble contrast techniques. For example, arteriovenous malformation (AVM) transit times which are crucial for proper glue embolization treatments, were shown to be about half when using droplets compared to those measured using soluble contrast techniques. In this work, factors such as x-ray pulse duration, frame rate, system spatial resolution (focal spot size), droplet size, droplet and system contrast parameters, and system noise are considered in relation to their affect on the accurate determination of droplet location and velocity.

Blood flow rate is an important parameter for functional evaluation of vascular disease. Instantaneous blood flow measurements from digital cerebral angiograms can be performed during endovascular interventional procedures providing interventional radiologists with minimally invasive real-time flow measurements. Distance-density curve matching (DDCM) methods are a promising class of videodensitometric techniques. However, published techniques have a relatively low theoretical maximum of measurable flow rate and sensitivity to noise and image artifacts. We investigate the use of alternative difference metrics along with curve fitting and extrapolation. These modifications can potentially reduce the influence of noise, image defects and flow irregularities. Extrapolation of difference profiles may overcome the theoretical limit for maximum measurable flow rate. The proposed methods were evaluated using both simulated angiograms and angiograms obtained by imaging a flow phantom under clinically realistic flow and contrast injection conditions. Our results indicate that under the conditions of constant flow the proposed modifications yield some improvement in both accuracy and reliability of instantaneous flow rate measurements. These improvements were the most noticeable during the early contrast wash-out phase, when the published DDCM methods were observed to fail.

Cerebral blood volume (CBV) is a major determinant of intracranial pressure (ICP). Hyperventilation is commonly employed to reduce raised ICP (e.g. in brain tumour patients) presumably through its effect on CBV. With the advent of slip- ring CT scanners, dynamic contrast-enhanced imaging allows for the measurement of CBV with high spatial resolution. Using a two-compartment model to characterize the distribution of X- ray contrast agent in the brain, we have developed a non- equilibrium CT method to measure CBV in normal and pathological regions. We used our method to investigate the effect of hyperventilation on CBV during propofol anaesthesia in rabbits with implanted brain tumours. Eight New Zealand White rabbits with implanted VX2 carcinoma brain tumours were studied. For each rabbit, regional CBV measurements were initially made at normocapnia (PaCO₂ 40 mmHg) and then at hyperventilation (PaCO₂ 25 mmHg) during propofol anaesthesia. The head was positioned such that a coronal image through the brain incorporated a significant cross-section of the brain tumour as well as a radial artery in a forelimb. Images at the rate of 1 per second were acquired for 2 minutes as Omnipaque 300 (1.5 ml/kg rabbit weight) was injected via a peripheral vein. In these CT images, regions of interest in the brain tissue (e.g. tumour, contra-lateral normal, and peri-tumoural) and the radial artery were drawn. For each region, the mean CT number in pre-contrast images was subtracted from the mean CT number in post-contrast images to produce either the tissue contrast concentration curve, or the arterial contrast concentration curve. Using our non- equilibrium analysis method based on a two-compartment model, regional CBV values were determined from the measured contrast concentration curves. From our study, the mean CBV values [+/- SD] in the tumour, peri-tumoural, and contra-lateral normal regions during normocapnia were: 5.47 plus or minus 1.97, 3.28 plus or minus 1.01, and 1.86 plus or minus 0.54 ml/100 g, respectively. Following hyperventilation, we found a significant decrease (p less than 0.025) of 10.4% in CBV in the peri-tumoural region, and no statistically significant change in CBV in the tumour or contra-lateral normal regions. We have developed a convenient method for measuring CBV in normal and pathological tissue using a slip-ring CT scanner. In a brain tumour model, we found that CBV was markedly increased in tumour and peri-tumoural regions compared to normal regions. Our results suggest that the reduction of raised ICP following hyperventilation during propofol anaesthesia may be mainly due to a reduction in CBV in the peri-tumoural tissue rather than in the bulk of the tumour or normal regions. Our method has the potential to provide further knowledge on the cerebral hemodynamics of space- occupying lesions during different anaesthetic interventions or treatment regiments.

The overriding incentive for accurate quantification of the functional status of children treated for brain tumors emerges from the clinician's desire to balance the efficacy and chronic toxicity of therapies used for the developing child. A hybrid combination of the Kohonen self-organizing map (SOM) for segmentation and a multilayer backpropagation (MLBP) neural network for classification removes observer variances to yield a reproducible and accurate identification of tissues. A group of 17 volunteers and 77 patients from a larger ongoing study of pediatric patients with brain tumors were used to investigate the sensitivity of segmented volumes to determine atrophy as measured by two radiologists. The atrophy study revealed a significant relationship for brain parenchyma, CSF and white matter volumes with atrophy while gray matter had no significant relationship. Brain parenchyma and subsequently white matter were found to be inversely proportional to increasing grades of atrophy. An additional study compared fifteen age-matched patients treated with irradiation and surgery with patients treated with surgery alone. The age-matched study of patients demonstrated that brain volumes in the irradiated patients were significantly decreased compared to those treated with surgery alone. Further investigation of this difference revealed that white matter was significantly reduced while gray matter was relatively unchanged.

Contrast-enhanced magnetic resonance (MR) imaging offers a minimally invasive method of investigating brain blood flow. This paper describes two different methods of extracting quantitative and qualitative information from this data. The first approach is to generate parametric images showing blood flow, blood volume and time-to-peak activity on a pixel by pixel basis. The second approach uses factor analysis. Principal components are extracted from the data and these orthogonal factors are then rotated to give a set of oblique factors, which satisfy certain simple constraints. In most cases three factors can be identified: a background or non- enhancing factor, an early vascular factor which is strongly correlated to arterial flow, and a late vascular factor which is strongly correlated to venous flow. The parametric and factor images are complimentary in nature: the former provides quantitative information that is readily understood by the clinician, while the latter makes no a priori assumptions about the underlying physiology and also allows more subtle changes in cerebral blood flow to be assessed. The factor images may also be of great value in defining regions of interest over which to carry out a more detailed quantitative analysis. This dual approach can be readily adapted to assess perfusion in other organs such as the heart or kidneys.

In this paper we present a method that characterizes a certain tissue class by the shape of the MR signal intensity versus time course obtained from the dynamic series images following a Gadolinium-DTPA bolus injection. This characterization is based on a pharmacokinetic model of the perfusion and leakage of contrast agent in the tissue. An objective classification of malignancy using a priori information is based on matching the actual enhancement time course of each pixel to reference time courses. Eigenimage filtering using characteristic time courses as feature vectors is proposed as an approach to reduce the dynamic series of images to a single image in which pixels with a close match to a particular feature are enhanced. This single image can be used as a mask to obtain homogenous regions for parameterization using the pharmacokinetic model. A newly developed algorithm enables the creation of training sets of standard time courses from dynamic series images of lesions with known histology in an objective way. An automatic segmentation of a new patient scan without user interaction is obtained using the training sets as feature vectors in the eigenimage filter.

Quantification of brain structure is important for evaluating changes in brain size with growth and aging and for characterizing neurodegeneration disorders. Previous quantification efforts using ex vivo techniques suffered considerable error due to shrinkage of the cerebrum after extraction from the skull, deformation of slices during sectioning, and numerous other factors. In vivo imaging studies of brain anatomy avoid these problems and allow repetitive studies following progression of brain structure changes due to disease or natural processes. We have developed a methodology for obtaining triangular mesh models of the cortical surface from MRI brain datasets. The cortex is segmented from nonbrain tissue using a 2D region-growing technique combined with occasional manual edits. Once segmented, thresholding and image morphological operations (erosions and openings) are used to expose the regions between adjacent surfaces in deep cortical folds. A 2D region- following procedure is then used to find a set of contours outlining the cortical boundary on each slice. The contours on all slices are tiled together to form a closed triangular mesh model approximating the cortical surface. This model can be used for calculation of cortical surface area and volume, as well as other parameters of interest. Except for the initial segmentation of the cortex from the skull, the technique is automatic and requires only modest computation time on modern workstations. Though the use of image data avoids many of the pitfalls of ex vivo and sectioning techniques, our MRI-based technique is still vulnerable to errors that may impact the accuracy of estimated brain structure parameters. Potential inaccuracies include segmentation errors due to incorrect thresholding, missed deep sulcal surfaces, falsely segmented holes due to image noise and surface tiling artifacts. The focus of this paper is the characterization of these errors and how they affect measurements of cortical surface area and volume.

The dynamic properties of human motor activities, such as those observed in the course of drawing simple geometric shapes, are considerably more complex and often more informative than the goal to be achieved; in this case a static line drawing. This paper demonstrates how these dynamic properties may be used to provide a means of assessing a patient's visuo-spatial ability -- an important component of neuropsychological testing. The work described here provides a quantitative assessment of visuo-spatial ability, whilst preserving the conventional test environment. Results will be presented for a clinical population of long-term haemodialysis patients and test population comprises three groups of children (1) 7-8 years, (2) 9-10 years and (3) 11-12 years, all of which have no known neurological dysfunction. Ten new dynamic measurements extracted from patient responses in conjunction with one static feature deduced from earlier work describe a patient's visuo-spatial ability in a quantitative manner with sensitivity not previously attainable. The dynamic feature measurements in isolation provide a unique means of tracking a patient's approach to motor activities and could prove useful in monitoring a child' visuo-motor development.

Nowadays, transgenic mice are a common tool to study brain abnormalities in neurological disorders. These studies usually rely on neuropathological examinations, which have a number of drawbacks, including the risk of artefacts introduced by fixation and dehydration procedures. Here we present 3D Fast Spin Echo Magnetic Resonance Imaging (MRI) in combination with 2D and 3D segmentation techniques as a powerful tool to study brain anatomy. We set up MRI of the brain in mouse models for the fragile X syndrome (FMR1 knockout) and Corpus callosum hypoplasia, mental Retardation, Adducted thumbs, Spastic paraplegia and Hydrocephalus (CRASH) syndrome (L1CAM knockout). Our major goal was to determine qualitative and quantitative differences in specific brain structures. MRI of the brain of fragile X and CRASH patients has revealed alterations in the size of specific brain structures, including the cerebellar vermis and the ventricular system. In the present MRI study of the brain from fragile X knockout mice, we have measured the size of the brain, cerebellum and 4^th ventricle, which were reported as abnormal in human fragile X patients, but found no evidence for altered brain regions in the mouse model. In CRASH syndrome, the most specific brain abnormalities are vermis hypoplasia and abnormalities of the ventricular system with some degree of hydrocephalus. With the MRI study of L1CAM knockout mice we found vermis hypoplasia, abnormalities of the ventricular system including dilatation of the lateral and the 4^th ventricles. These subtle abnormalities were not detected upon standard neuropathological examination. Here we proved that this sensitive MRI technique allows to measure small differences which can not always be detected by means of pathology.

Approach reported in this paper uses a sliding (rectangular) window on the original fMRI scans to produce the new scans. By selecting parameters of sliding windows, autocorrelation function (acf) among new scans can be reduced to a predicted level, and fMRI analysis will be performed on these new scans. Theoretical analysis showed that this level is 1/k (k is the window size). Simulations by using different window structures verified this conclusion and the graphic illustration provides an intuitive explanation.

A novel Local Principal Component Analysis (LPCA) technique is presented in this paper for activation detection in functional Magnetic Resonance Imaging (fMRI) without explicit knowledge about the shape of the activation signal. The proposed LPCA method is very different from the traditional PCA methods for fMRI signal detection in principle. At first, our LPCA algorithm does not require any orthogonality assumption between the activation signal and other signal components, while the traditional PCA methods are based on this assumption. In addition, our LPCA algorithm applies PCA to the temporal sequence of each individual voxel instead of applying PCA to the whole data set. In our algorithm, we first apply a linear regression procedure to alleviate the common baseline drift artifact. Then the baseline-corrected temporal signals are partitioned into active and inactive segments according to the paradigm used for the fMRI data acquisition. Several most dominant principal components are computed from all these segments for each voxel by the PCA. By projecting the segments of each voxel onto a linear subspace formed by the corresponding dominant principal components, two separate clusters are formed from the active and inactive segments. An activation measure is defined based on the degree of separation between these two clusters in the projection space. Experimental results of applying our LPCA algorithm to detect fMRI activation signals on various data sets are given. From our experiments, the LPCA algorithm in general provides 4 - 6 times signal-to-noise ratio (SNR) improvement over the standard t-test method.

Phase-contrast (PC) method of magnetic resonance imaging (MRI) has bee used for quantitative measurements of flow velocity and volume flow rate. It is a noninvasive technique which provides an accurate two-dimensional velocity image. Moreover, Phase Contrast Cine magnetic resonance imaging combines the flow dependent contrast of PC-MRI with the ability of cardiac cine imaging to produce images throughout the cardiac cycle. However, the accuracy of the data acquired from the single through-plane velocity encoding can be reduced by the effect of flow direction, because in many practical cases flow directions are not uniform throughout the whole region of interest. In this study, we present dynamic three-dimensional velocity vector mapping method using PC-MRI which can visualize the complex flow pattern through 3D volume rendered images displayed dynamically. The direction of velocity mapping can be selected along any three orthogonal axes. By vector summation, the three maps can be combined to form a velocity vector map that determines the velocity regardless of the flow direction. At the same time, Cine method is used to observe the dynamic change of flow. We performed a phantom study to evaluate the accuracy of the suggested PC-MRI in continuous and pulsatile flow measurement. Pulsatile flow wave form is generated by the ventricular assistant device (VAD), HEMO-PULSA (Biomedlab, Seoul, Korea). We varied flow velocity, pulsatile flow wave form, and pulsing rate. The PC-MRI-derived velocities were compared with Doppler-derived results. The velocities of the two measurements showed a significant linear correlation. Dynamic three-dimensional velocity vector mapping was carried out for two cases. First, we applied to the flow analysis around the artificial heart valve in a flat phantom. We could observe the flow pattern around the valve through the 3-dimensional cine image. Next, it is applied to the complex flow inside the polymer sac that is used as ventricle in totally implantable artificial heart (TAH). As a result we could observe the flow pattern around the valves of the sac, though complex flow can not be detected correctly in the conventional phase contrast method. In addition, we could calculate the cardiac output from TAH sac by quantitative measurement of the volume of flow across the outlet valve.

Deformation and motion of the Mitral Annulus (MA) is closely related to the left ventricular function. Measurement and visualization of the characteristic parameters in 3D will help in understanding the relationship. Data for this study was acquired from patients undergoing transesophageal echocardiographic examination with the transducer aligned along the axis roughly perpendicular to the annuli, and rotated automatically to cover 360 degrees. ECG gated images were acquired at 24 angles for each phase of the cardiac cycle. The annuli hinge points were identified from each echo image and the annuli reconstructed. The parameters measured to characterize the annuli were: (1) area of projection, (2) non- planarity, (3) excursion of annulus centroid, (4) change in the annulus orientation. We validated the method using a wire loop shaped in the form of a saddle and a planar rubber ring imaged in a water bath at different orientations. Four MAs were reconstructed using this method. Two were patients with dilated cardiomyopathy (DCM) and two were patients with normal ventricular function. The change in parameters was measured from systole to diastole. Percentage change in area (29% vs. 16%) and excursion (8 mm vs. 3 mm) were much larger for normals than for patients. While, changes in non-planarity (20%) and orientation (6 deg) were similar. These preliminary results show that MA parameters do reflect the abnormality, and could be used for diagnosis and prognosis of patients with bad ventricles.

We applied the spectral imaging technique to the remote sensing of physiological responses on the human body. Blood, sweat and thermal distributions and their fluctuation are important and useful information to estimate the physiological state or thermal comfort of a person. Such information can be obtained as images by using cameras which can detect different wavelength regions. The blood distribution can be observed over a 430 nanometer wavelength region by the absorption pattern of oxidized hemoglobin contained in blood. Also information of sweat distribution can be obtained over a 1.9 micrometer region by the absorption pattern of water. Thermal cameras can acquire a thermal distribution of the human body without contact. We therefore attempt to observe simultaneously blood, sweat and thermal distributions and their fluctuations by spectral imaging. Some experimental results are shown. The sensitivity of this technique is discussed.

In addition to conventional MR parameters such as proton density and relaxation times T1 and T2, further specific parameters (magnetization transfer, apparent diffusion, temperature, blood oxygen level) have been observed by recent developments in Magnetic Resonance Imaging. When employing several approaches, such as indicator dilution theory or pharmacokinetic modeling, further valuable information can be derived from MR raw data sets. This additional information could give more insight into various clinical areas. It can be utilized to investigate hemodynamic perturbations and tumor- related pharmacokinetic features. Most of these techniques rely on a large amount of MR raw data. Hence, an efficient strategy for data organization and postprocessing and clear image data presentation is necessary for these techniques to be implemented as clinical routine work. A novel software concept has been developed to address all these issues. It is based on a more general framework of basic classes for data IO and manipulation, viewing and image processing. For a maximum of flexibility, the presented design separates data management and image presentation from dedicated MR analysis. It provides facilities for incorporation of user-defined routines and a comprehensive set of basic functions which enable rapid software prototyping.

This paper is a review of our recent studies using a texture- based tissue characterization method called the Adaptive Multiple Feature Method. This computerized method is automated and performs tissue classification based upon the training acquired on a set of representative examples. The AMFM has been applied to several different discrimination tasks including normal subjects, subjects with interstitial lung disease, smokers, asbestos-exposed subjects, and subjects with cystic fibrosis. The AMFM has also been applied to data acquired using different scanners and scanning protocols. The AMFM has shown to be successful and better than other existing techniques in discriminating the tissues under consideration. We demonstrate that the AMFM is considerably more sensitive and specific in characterizing the lung, especially in the presence of mixed pathology, as compared to more commonly used methods. Evidence is presented suggesting that the AMFM is highly sensitive to some of the earliest disease processes.