Dynamic Spatial Reconstructor Research Articles

High-speed 3D imaging of the beating heart using temporal estimation

CT scanners require a finite time to collect projections around the object which is to be reconstructed. In order to obtain high temporal resolution images of moving structures, reconstruction from relatively few views must be carried out, since the extended time which would be required to collect a large number of views causes blurring in the reconstructions. This paper describes a method which improves the quality of images reconstructed from a relatively small number of measured views. This is accomplished by providing the reconstruction algorithm with additional unmeasured projections which are estimated in the time domain with the aid of Kalman-type filters. The procedure is tested using mathematical simulations and from data obtained by scanning a dog in the Dynamic Spatial Reconstructor—a high temporal resolution, synchronous volume (3-dimensional), computed tomography scanner.

Heart-lung interaction: effect on regional lung air content and total heart volume.

To study the interactions between and within the heart and lungs, end-diastolic (ED) and end-systolic (ES) volumes and intrathoracic location of the heart, and the regional air content, volume and geometry of the lungs, were measured from three-dimensional image data generated with the Dynamic Spatial Reconstructor (DSR). The DSR was used to scan the full thoracic extent of anesthetized dogs and sloths at selected transpulmonary pressures. The results show that the dependent to nondependent gradient of regional lung opacity (or conversely regional air content) in the supine animal was not present in the prone animals. While the rib cage and diaphragm of the dog deformed markedly, the shape of the sloth's rib cage and diaphragm remained essentially constant with change in body orientation. As a consequence of these findings, we deduce that the observed change in gradient of regional lung air content in both dog and sloth are in response to changes in the intrathoracic position of the heart which alter ventral lung geometry and not a response to changes in rib cage or diaphragm geometry. In a second series of studies we reconstructed the 3-D extent of the heart at ED and ES in supine anesthetized dogs and demonstrated that the total heart volume (THV) (i.e. contained by the pericardial sac) during sinus rhythm differs by less than 5% between ED and ES. The DSR image data show that this is achieved by the epicardial apex remaining essentially fixed and that the plane containing the atrio-ventricular valves moves like a plunger towards the apex in systole. When atrial fibrillation is present, the THV no longer remains constant and decreases during systole, presumably because of increased stiffness of the atrial myocardium. We conclude from the experimental results that the heart plays an important role in determining regional differences in alveolar expansion, and that by maintaining a constant THV, the heart minimizes energy expenditure which would be caused by moving the lung.

Quantification of thoracic volumes by three-dimensional imaging.

End-expiratory thoracic cavity volume (Vthx) was measured in eight volunteers lying supine by three-dimensional X-ray computed tomography using the Dynamic Spatial Reconstructor. Untrapped end-expiratory pulmonary gas volume at functional residual capacity (FRC) was determined by nitrogen clearance. Both measurements were done before and after induction of anesthesia-paralysis. After induction of anesthesia-paralysis, Vthx and FRC were consistently and significantly (P less than 0.01) reduced by 0.28 +/- 0.22 (SD) and 0.59 +/- 0.24 liter, respectively. The reduction of FRC was larger than the reduction of Vthx (delta Vthx) in six of the eight subjects, a finding suggesting that intrathoracic fluid (blood) plus trapped gas volume (Vtt) increased. Changes in Vthx were partitioned into volume changes from the thoracic rib cage (delta Vrc) and from shape and/or position changes of the diaphragm (delta Vdi). delta Vrc contributed significantly (0.17 +/- 0.15 liter, P less than 0.02) to delta Vthx, whereas delta Vdi contributed only in four of the eight subjects. We conclude that delta Vrc, delta Vdi, and delta Vtt contribute to the reduction of FRC after induction of anesthesia-paralysis in humans; the relative contribution of them varies among subjects.

Measurement of three-dimensional anatomy and function of pulmonary arteries with high-speed x-ray computed tomography.

Pulmonary arterial geometry was quantitatively evaluated in five dogs. One was a control dog, two were subjected to monocrotaline administration, and two had a subclavian-to-pulmonary artery shunt. Measurements of pulmonary arterial anatomy were made with the dynamic spatial reconstructor (DSR), a high-temporal-resolution, volumetric, roentgenographic, computed tomographic scanner. In the control dog, pulmonary arterial cross-sectional areas and segment lengths were measured both from images generated from DSR scans obtained during injection of contrast medium and from a methylmethacrylate cast of the same vascular tree. The correlations between the DSR and cast-based measurements of cross-sectional area and segment lengths were 0.98 and 0.97, respectively. Over periods of eight and 11 months' observation, the pulmonary arterial pressure in the two dogs with surgically created left-to-right shunts increased from 29/15 to 42/17 mm Hg and 19/12 to 25/16 mm Hg, respectively. The cross-sectional areas increased progressively in the proximal portion of the main lobar arteries, remained unchanged in the midportion, and decreased in the distal portion of these arteries. Similar changes occurred in one of the two dogs treated with monocrotaline injections, while in the other the taper of the distal vessels did not change significantly. Pulmonary vascular resistance, computed from cardiac output and pulmonary artery pressure, increased proportionately to the decrease of the cross-sectional areas of distal vessel in both the monocrotaline administration dogs; the resistance decreased proportionately to the increase of the pulmonary arterial cross-sectional areas in both of the surgical dogs.

Angiography for Delineation of Systemic-to-Pulmonary Shunts in Congenital Pulmonary Atresia: Evaluation With the Dynamic Spatial Reconstructor

In patients with pulmonary valve atresia, the arterial supply to the lungs is derived entirely from the aorta or its major branches. The anatomic detail of this arterial supply is usually identified by obtaining multiple selective arteriograms involving numerous injections of contrast material and x-ray exposures. In this study, we demonstrated that the information obtained from conventional diagnostic angiography in two patients with congenital pulmonary atresia was also clearly evident on scans from the dynamic spatial reconstructor (DSR), a high-speed three-dimensional computed tomographic system. Thus, numerous shunts, stenoses, and systemic collateral arteries were demonstrated by DSR scanning, which involved only one or two nonselective injections of a contrast agent and less x-ray exposure than that normally necessitated by angiographic study.

3-D Reconstruction of the Heart from Few Projections: A Practical Implementation of the McKinnon-Bates Algorithm

The implementation of a method for improving the quality (by reducing artifacts) of images reconstructed from small numbers of projections is described. The technique is based on an algorithm originally proposed by McKinnon and Bates [1] and includes enhancements suggested by Heffernan and Robb [2]. Emphasis is placed on practical aspects of algorithm implementation, including important noise-suppressing refinements not previously incorporated into the algorithm. The efficacy of the method is demonstrated using actual X-ray projection data of the intact beating heart recorded with the Mayo Clinic's dynamic spatial reconstructor. The implications of improved "stop-action" reconstruction of the heart are considered in the context of additional objectives, including automatic object recognition and definition, and improved four-dimensional reconstruction of the beating heart.

Regional myocardial volume perfused by the coronary artery branch: estimation in vivo.

Seven closed-chest dogs were anesthetized with intramuscular Innovar and a N2O/O2 gas mixture. Maximal coronary vasodilation was induced by intra-aortic injection of nitroglycerin (200 micrograms/kg) and continuous infusion of adenosine (1 mg/kg/min) into the main pulmonary artery. Superselective coronary arteriograms were obtained at varying distances along the left circumflex and left anterior descending coronary arteries during scans using the dynamic spatial reconstructor rapid tomographic x-ray scanner. At end-diastole the images of the coronary arteries and opacified myocardium were analyzed for cross-sectional area (CSA) of the coronary artery lumen and regional myocardial volume (Vdsr) perfused by that coronary artery. Postmortem regional myocardial volume (Vpath) was related to the volume perfused by the same artery measured in vivo by the dynamic spatial reconstructor as follows: Vdsr = 4.56 + 0.93 Vpath (r = .98, p less than .001). In addition, the cross-sectional area of the coronary artery supplying a volume of myocardium was related to that volume as follows: Vdsr = - 1.95 + 6.34 CSAmax (r = .88, p less than .001). This suggests that a maximally dilated coronary artery luminal cross-sectional area is linearly related to the volume of muscle it perfuses. We speculate that this relationship may be useful in expressing the physiologic significance of coronary arterial narrowing.

Anatomic and kinematic analysis of the human forearm using high-speed computed tomography.

A new technique using computerised tomographic images was investi gated that allows a noninvasive description of the anatomy and kinematics of the human musculoskeletal system. These images were generated by the Dynamic Spatial Reconstructor, a high-speed CT scanner, with the ability to simultaneously scan multiple thin slices. Pronation/supination in a cadaveric forearn was used as the model for investigation. Three-dimensional shaded surface displays of the bony forearm were generated. These images are dimensionally accurate and allow the superposition of the results of mathematical modelling on the image for visual examination. The spatial location of bony markers was measured and calculated screw axis data superimposed on a computer generated surface display of the forearm skeleton. Applications of the technique include anatomic measurements not easily made with other systems, such as radial-ulnar interosseous distance as a function of forearm rotation and length and the amount of forearm rotation possible at various levels of forearm lengths.

Computer-aided analysis of regional lung air content using three-dimensional computed tomographic images and multinomial models

Fast multisliced X -ray computed tomography yields unique anatomic and physiologic information regarding the function of the heart, lungs, and circulation within the negative pressure environment of the intact thorax. We have applied multinomial models to the distribution of reconstructed lung air content within the large data volumes (up to 4 million 16-bit voxels) generated by the Dynamic Spatial Reconstructor to aid in data analysis. Such modeling coupled with easily accessible user interactive computer graphics routines has proven useful in (1) smoothing over discrete regions of artifact in reconstructed images; (2) allowing for the streamlining of previously time intensive analysis tasks; and (3) providing a convenient numerical framework from which parameters determining regional lung expansion patterns can be assessed.

A Method for Quantitative Display of Three-Dimensional Regional Myocardial Function

The DSR (dynamic spatial reconstructor), a multiple X-ray source scanner that generates stop action three-dimensional (3-D) images of a cylindrical volume, was used for quantitative imaging of left ventricular 3-D wall geometry and function in experimentally induced canine left ventricular myocardial infarction. Impaired regional myocardial function was induced by myocardial ischemia or infarction in four mongrel dogs by closed-chest occlusion of the proximal left anterior descending (LAD) coronary artery. At intervals of 6-14 weeks post occlusion, the dogs were scanned with the DSR during biatrial contrast injection. The 3-D shape, extent, and function of hypokinetic myocardium was measured from the DSR images utilizing measurement of the rate of local systolic wall thickening to detect regions of normal, ischemic, or scarred myocardium. The results were compared to scar size and anatomic distribution measured at postmortem examination. The anatomic extent and relationship of hypocontractile to normally contracting muscle was visualized by computer generated, pseudo 3-D shaded surface displays of the left ventricular chamber and by topographic projections of regional wall thickening rates onto a map of the left ventricular endocardial surface. The location of myocardial infarction and the surrounding zone of impaired function is clearly defined by this 3-D CT scanning procedure. The display method presented here provides both localization and quantification of the volume of ischemic and infarcted myocardium.

Invariant total heart volume in the intact thorax

Fast multisliced computerized tomography (Dynamic Spatial Reconstructor or DSR) was used to study the change in total heart volume (content of the pericardial sac) between end diastole (ED) and end systole (ES) with lungs held at 0 and 15 cmH2O airway inflation pressure (Paw). Nine dogs were anesthetized and scanned in the DSR. At 0 cmH2O airway pressure, mean total heart volume changed, on the average, only 2.7 +/- 0.6% (SE) of its ED volume comparing ED with ES. With lungs inflated to 15 cmH2O Paw, total heart volume decreased 12 +/- 0.5%. However, at this new heart volume, the change in total heart volume between ED and ES was only 1.7 +/- 0.5%. The data indicate that the epicardial apex of the heart remains relatively stationary while the atrioventricular groove moves toward the apex during systole and away from the apex during diastole. Thus the atria and ventricles empty and fill reciprocally even when the pericardial contents do not fully distend the pericardial sac. The invariant total heart volume observed in these species would minimize the work of the heart by maximizing the percentage of work expended to move blood and minimizing the work expended to move tissue (i.e., lung).

Effect of body orientation on regional lung expansion: a computed tomographic approach.

The Dynamic Spatial Reconstructor (DSR) was used to study in vivo lung geometry and function. By replacing the lungs of three dogs with potato flakes and ping-pong balls of known air content and scanning these realistic phantoms in the DSR we have estimated accuracy of lung density to be within 7% and have demonstrated a high (+/- 3%) internal consistency (relative density within dogs). Change in total lung air content (y) as calculated from DSR volume imaging of anesthetized dogs matched the known inflation steps (x) to within 7% [range was 1-7% with a mean of 3 +/- 0.5% (SE)]. A gradient of decreasing percent lung air content was measured in the ventral-dorsal direction at functional residual capacity (FRC) in the supine body posture (y = 3.29% air content/cm lung height + 46.48% air content; r = 0.90). Regional lung air content change with lung inflation was greatest in the dependent lung regions. In contrast, regional lung air content at FRC was approximately uniform along the ventral-dorsal direction with the dog in the prone posture and was 66 +/- 0.6% (SE). Ventral-dorsal gradients in lung air content measured within an isogravimetric plane of the dogs in the left or right lateral body posture suggest that regional differences in lung air content cannot be explained solely on the basis of a direct gravitational effect on the lung. Evidence is presented to suggest a possible major role of the intrathoracic position of the mediastinal contents in determining these lung air content distributions.

Effect of body orientation on regional lung expansion in dog and sloth.

Recent studies (E.A. Hoffman, J. Appl. Physiol. 59: 468-480, 1985) using fast multisliced X-ray computed tomography have demonstrated a ventral-dorsal gradient of fractional lung air content (3.29% air/cm lung height) in supine dogs and an essentially uniform ventral-dorsal air content distribution in the prone dogs [mean = 66 +/- 0.6% (SE) air content]. Since the prone orientation is the dog's normal body posture, we sought to study an animal whose normal body posture was "opposite" to that of the dog. Four two-toed sloths were scanned in the Dynamic Spatial Reconstructor in the prone and supine postures. A supine fractional air content gradient was demonstrated with a regression equation of y = 2.09x + 74.3 (r = 0.92), where y is percent air content and x is vertical height in the lung, and ventral-dorsal air content distribution in the prone posture was uniform with a mean of 85 +/- 0.4% (SE) air content. The low functional residual capacity lung density in the sloth was attributable to unusually large alveoli. The mean heart volume-to-body weight ratio in the dogs was 16.4 +/- 0.6 (SE) ml/kg and that in the sloth was 7.3 +/- 0.4 (SE) ml/kg. Mean lung volume-to-body weight ratios for dogs and sloths were 57 +/- 7 (SE) and 89 +/- 6 ml/kg, respectively. Of particular interest was the fact that large changes in prone vs. supine rib cage and diaphragm geometry previously found in dogs did not occur in sloths, though significant alterations of ventral and dorsal lung geometry prone vs. supine were demonstrated, and lung shape changes in both dog and sloth are attributable to shifts in the intrathoracic position of mediastinal structures.

Shape and dimensions of cardiac chambers: importance of CT section thickness and orientation.

Three-dimensional (3D) computed tomography (CT) scan data were used to quantitate the geometry of all heart chambers. The Dynamic Spatial Reconstructor (DSR) was used to scan dogs with in situ casts of the cardiac chambers. Chamber volumes estimated from DSR images were accurate within 5% of water displacement volume measurements of the actual casts for chambers greater than 11 ml and within 10% of water displacement volumes for chambers less than 11 ml. Anatomic features of the actual cast correlated closely with anatomy visible in computer-generated surface images of the 3D DSR image data. The important effect of reconstructed section thickness and orientation on the fidelity of 3D cardiac geometry is demonstrated.

Three-Dimensional Cardiac Anatomy and Function in Heart Disease in Adults: Initial Results With the Dynamic Spatial Reconstructor

The dynamic spatial reconstructor, or DSR, is a unique high-speed volume-imaging x-ray scanner based on computed tomographic principles. In this report, we present data obtained from the first feasibility DSR studies of adult patients with heart disease. Information from three patients--one with hypertrophic obstructive cardiomyopathy, one with calcific aortic valvular disease, and one with a left ventricular aneurysm--is described in detail. The mean DSR scanning time for each patient was 20 seconds, and the mean total irradiation to the sternum was 15.3 R. Transverse cross sections were reconstructed and then retrospectively reformatted to provide operator-selected oblique sections in space (for example, long-axis and short-axis sections of the left ventricle), to follow these sections through time (such as from end-diastole through end-systole), and to create three-dimensional displays (for instance, of the left ventricular chamber). Unique quantitative measurements of structure and function were made by using these images. For generation of most imaging data, only one injection of contrast material into the right side of the heart is necessary. Clinically useful three-dimensional dynamic imaging data can be acquired from adult patients with heart disease by using the DSR. Compared with conventional angiocardiography, DSR studies can provide information with less x-ray exposure and fewer angiographic injections.

Renal cortical perfusion--preliminary experience with the dynamic spatial reconstructor (DSR).

The three-dimensional image data generated by the Dynamic Spatial Reconstructor (DSR) enables measurement of the three-dimensional distribution of blood supply in organs. We have applied this imaging technique to evaluate renal cortical blood flow distribution and compare it with distribution of radiolabeled microspheres. The DSR, a high temporal resolution volumetric roentgenographic computed tomographic scanner, was used to scan the volume containing a kidney in 0.13-0.26 s and repeating this scan 8-4 times per s for six s during a renal arteriogram. Five anesthetized dogs were studied in the prone position with the left kidney exteriorized through a flank incision. An electromagnetic flowmeter was placed around the renal artery and a needle placed retrograde into the artery for injection of a 2 cc bolus of contrast agent. During the scan the contrast agent was injected over a four s period during which radioactively labelled microspheres were injected into the left atrium. The tomographic images of approximately 10 parallel, 5 mm thick sagittal slices corresponding to the slices of the kidney used for counting microspheres in the cortical layers were displayed and analyzed. The time point chosen for analysis was the one in which peak brightness (i.e., concentration of contrast agent) was detected in the cortex. The spatial distribution of peak brightness values was compared to the number of microspheres at the same sampling locations. The microsphere-based value of regional cortical blood flow fell below the regression line for the juxtamedullary cortex and above for the outer cortex. This result is consistent with the preferential distribution of microspheres to the outer cortex whereas the contrast agent distributed more uniformly throughout the cortex.

Evaluation of coronary anatomy using high-speed volumetric computed tomographic scanning

Although coronary angiography involves minimal risk, it is sufficiently invasive that symptoms suggestive of myocardial ischemia are usually needed before the procedure is indicated, and presymptomatic patients in an at-risk group generally do not undergo catheterization. A practical method for obtaining coronary anatomy is computed tomographic (CT) scanning of contrast-filled coronary arteries. To visualize the epicardial coronary branches, a 3-dimensional (volume) CT image must be generated that contains the entire coronary tree, and the volume must be scanned fast enough to minimize image blurring and distortion due to cardiac motion. These features are available with the dynamic spatial reconstructor (DSR). In this report, we describe our initial use of the DSR to evaluate coronary anatomy from a single selective contrast medium injection into the left coronary artery. The DSR is a prototype instrument (Fig. 1) capable of scanning a 3-dimensional volume 21.5 cm in length and in diameter in 0.01 second and at a rate of 60 times/s. It contains 14 x-ray sources with 14 corresponding video based imaging chains. The x-ray tubes and image chains are mounted on a rotating turret (15 rpm) and are pulsed sequentially to provide 12-degree increments in viewing angle around the semicircular scanning arc. Detailed description of the DSR has been provided in previous studies. 1–3

Anatomy and function of the heart and intrathoracic vessels in congenital heart disease: Evaluation with the dynamic spatial reconstructor

The Dynamic Spatial Reconstructor is a unique high speed volume imaging X-ray scanner based on computed tomographic principles. It has several potential advantages over conventional angiographic methods, including reduced invasion, reduced rate of false negative results and increased accuracy of measurements of structure and function. To evaluate the utility of the Dynamic Spatial Reconstructor in the investigation of congenital heart disease, scanning was performed in several pediatric patients with pulmonary valve atresia. Early results show that three-dimensional images of the cardiac chambers and intrathoracic vessels can be acquired and displayed with the system. All the information necessary for quantitative analysis of the pulmonary arteries can be obtained with a single scan involving injection of 0.6 to 2.0 cc/kg of contrast agent and radiation exposure of 0.54 to 0.95 rad/s over 5 to 8 seconds.

Three-dimensional dynamic x-ray-computed tomography imaging of stomach motility.

The dynamic spatial reconstructor (DSR), an x-ray-computed, tomography-based imaging machine, scans a cylindrical volume up to 39 cm in transaxial diameter and 21.5 cm in axial height with scan repetition rate up to 60 per second. We applied this technique to three dogs in order to investigate the accuracy with which stomach shape, dimensions, and motility can be measured. Single peristaltic waves of one dog's stomach filled with air and two other dogs' stomachs filled with different volumes of contrast-enhanced semisolid meals (10% Gastrografin) were scanned at 2-s intervals at scan apertures between 0.011- and 1-s duration. On completion of the scan sequence, images of oblique sections perpendicular to the local long axis of the stomach fundus, corpus, and antrum were computed from the DSR scan data. Measurements of differential motion of defined points on the stomach wall surface, propagation velocities, acceleration, and amplitudes of peristalsis as well as gastric volume determinations were made. This approach overcomes problems of superposition and ambiguity inherent in projection imaging. Measurements of the volumes of ingested first meal were accurate to within 1.5% (of the actual volume of solid food introduced into the stomach) using a 0.40-s scan aperture. Wall motion could be detected with an SEE of 1.5 mm (pixel size used was 1.4 mm) using a 0.40-s scan aperture.

Coronary angiographic examination with the dynamic spatial reconstructor.

The dynamic spatial reconstructor (DSR), a high temporal resolution, three-dimensional roentgenographic, computed tomography scanner, was used to scan the coronary arteries of five dogs. After one injection of contrast medium into the left main coronary artery or into the aortic root of each dog, all major epicardial coronary arteries and the septal artery could be imaged. After selective removal of all nonangiographically enhanced anatomical structures within the three-dimensional images, pseudo three-dimensional displays were generated to show the arteries from all possible views--even the strictly cranial view--so that superposition of arteries could be avoided. The geometric accuracy of the images was evaluated by comparing coronary arterial segment length measured by the DSR with postmortem measurements (r = .99, SEE = 3.12 mm, n = 35) and by measuring the stenosis produced by nine hollow plastic cylinders lodged in coronary arterial lumina via percutaneous catheterization. Three of the plugs had irregular noncircular lumina drilled into them so that the percentage of stenosis seen in some projection images underestimated the severity of stenosis. The DSR overestimated the length of a 3 mm long stenosis by 0.4 +/- 0.5 mm. The percent area reduction caused by the hollow cylinders varied between 53% and 92% and was underestimated by the DSR on the average by 7%. The correlation between DSR measurements and the direct measurements was r = .85 (SEE = 5%, n = 9). The volume of plastic in each plug (average 13.2 mm3), calculated from the length, cross-sectional area, and degree of the stenosis, showed a better correlation (r = .90, SEE = 2 mm3) with the actual plastic volume.(ABSTRACT TRUNCATED AT 250 WORDS)