Paul L. Lee

Evaluation of computerized edge tracking for quantifying intima-media thickness of the common carotid artery from B-mode ultrasound images

A new method to measure carotid intima-media thickness (IMT) from B-mode ultrasound images was developed that utilizes automatic tracking of the lumen-intima and media-adventitia echoes. Phantom studies and human replicate studies under typical clinical protocols for common carotid IMT measurement were carried out to assist in evaluation of the method. A lucite step wedge phantom was used to show that incorporation of sub-pixel interpolation to locate echo boundaries allowed detection of changes in the echo separation that were 5-10 times smaller than the axial resolution of the ultrasound transducer. For average IMT measured in the distal common carotid artery (CCA) wall in 24 subjects scanned twice within 60 days, mean absolute difference was 0.036 mm with a standard deviation of 0.045 mm. Replicate scans obtained 1 week apart of eight subjects by three sonographers showed the intersonographer variability was 5.4%. In another study of 12 subjects scanned every 4 months for 48 months, the root mean square deviation of the IMT measurements from a linear regression line was 0.030 mm. These data indicate that the method is equally precise over short intervals (60 days) and over long intervals (48 months). The new automated computerized edge tracking method presented in this paper represents an advance for image analysis of B-mode ultrasound images of common carotid IMT with measurement variability substantially reduced (2 to 4 times) compared with currently available manual methods.

Improved common carotid elasticity and intima-media thickness measurements from computer analysis of sequential ultrasound frames

B-mode ultrasound has gained popularity as a non-invasive method for direct visualization of superficial vessels. With B-mode ultrasound, arterial stiffness can be directly measured since image acquisition of the arterial wall thickness and vessel diameter can be obtained simultaneously in a dynamic fashion throughout the cardiac cycle. Recently, a method was developed to measure carotid arterial diameter and intima-media thickness (IMT) from B-mode images that utilizes computerized edge tracking-multiframe image processing that automatically measures arterial diameter and IMT in multiple sequential frames spanning several cardiac cycles. To evaluate this method, replicate B-mode common carotid artery ultrasound examinations and blood pressure measurements were obtained in 24 subjects 1-2 weeks apart. Approximately 80 sequential frames spanning two cardiac cycles were analyzed from each ultrasound examination to obtain maximum arterial diameter (D(max)), minimum arterial diameter (D(min)), and IMT using a computerized edge detection method. The intraclass correlations of D(max), D(min), and IMT were 0.97-0.99 and the mean absolute difference for these measurements were 0.03-0.11 mm. The coefficient of variation for D(max) and D(min) were 1.28 and 1.18%, respectively. The intraclass correlation for several standard arterial stiffness indices, Petersons elastic modulus, Youngs modulus, arterial distensibility, compliance, and the beta stiffness index ranged between 0.84 and 0.89. Additionally, it was determined that averaging IMT over five frames centered at D(min) reduced single frame IMT measurement variability by 27% (P=0.005) compared with IMT measured from a single frame corresponding to D(min). Comparison of the phasic relationship of D(max) and D(min) measured from the B-mode ultrasound image with the simultaneously recorded electrocardiogram (ECG) signal in the 24 subjects, provided a more accurate method of frame selection for arterial diameter extrema independent of the ECG signal. The method of computerized edge detection-sequential multiframe image processing presented in this paper represents a technological advance for image analysis of B-mode ultrasound images of common carotid arterial dimensions that is highly reproducible and directly applicable to noninvasive imaging of atherosclerosis.

Hidden Markov model analysis of force/torque information in telemanipulation

A new model is developed for prediction and analysis of sensor information recorded during robotic performance of tasks by telemanipulation. The model uses the Hidden Markov Model (stochastic functions of Markov nets; HMM) to describe the task structure, the operator or intelligent controllers goal structure, and the sensor sig nals such as forces and torques arising from interaction with the environment. The Markov process portion encodes the task sequence/subgoal structure, and the observation densities associated with each subgoal state encode the expected sensor signals associated with carry ing out that subgoal. Methodology is described for con struction of the model parameters based on engineering knowledge of the task. The Viterbi algorithm is used for model based analysis of force signals measured during experimental teleoperation and achieves excellent segmen tation of the data into subgoal phases. The Baum-Welch algorithm is used to identify the most likely HMM from a given experiment. The HMM achieves a structured, knowledge-based model with explicit uncertainties and mature, optimal identification algorithms.

Telerobot Configuration Editor

The Telerobot Configuration Editor (TCE), an iconic graphical user interface based on the X Windows environment, is described. TCE offers an easy way to configure a complex telerobotic system which requires numerous parameter settings to access the full capabilities of the system. The screen layout design of TCE, which utilizes various visual coding mechanisms to make the hierarchical structure of the TCE visible to the user, is examined in detail. Two key features of the design-the macro configuration buttons and the system access levels-are examined. The macro configuration buttons can store the full or a partial set of configuration parameters, and these macros can be used as high-level configuration parameters that are well suited to the physical tasks such as peg insertion task. The system access levels limit the user access to the system, based on ability and need. Other design features are described.<<ETX>>

Multi-dimensional hidden Markov model of telemanipulation tasks with varying outcomes

Three multi-step teleoperation tasks were successfully modeled with a hidden Markov model (HMM). Tasks that would proceed through different paths as determined by an event either internal or external to the task were designed. The tasks can be described by a state transition diagram containing a fork through which two alternative task outcomes can be followed. The model was used to identify correctly the sequence of task progression from the recorded sensor data. Previous work with HMMs was extended by generalizing the model to encompass multi-dimensional sensor signals consisting of a mix of force, torque, and position signals. The addition of multi-dimensional sensor information significantly improved the ability of the Viterbi decoding algorithm to identify the series of events.<<ETX>>

Compensatory Vascular Changes of Remote Coronary Segments in Response to Lesion Progression as Observed by Sequential Angiography From a Controlled Clinical Trial

BACKGROUND Local coronary artery enlargement to compensate for atherosclerotic plaques preserves the vessel lumen. The extent to which coronary segments remote from progressing lesions enlarge is unknown. This is clinically relevant since compensatory enlargement may be important in determining whether clinical complications result from progression of coronary artery disease (CAD). Additionally, compensatory change has implications for quantitative coronary angiographic (QCA) trials, since the effect of progression on diameter means may be mitigated by compensatory changes in remote coronary segments when QCA change is averaged over all lesions. METHODS AND RESULTS Serial QCA data from 78 subjects in the Monitored Atherosclerosis Regression Study were used to demonstrate compensatory changes in coronary segments remote from progressing or regressing lesions. Coronary segments were first classified as progressing (regressing) if percent diameter stenosis (PS) increased or decreased by > 10 with a concurrent decrease or increase in minimum lumen diameter (MLD) of either > 0.32 mm or > 10% of the normal baseline reference diameter (DNORM). Segments not meeting these criteria were labeled stenosis stable. Stenosis-stable segments opposite progressing lesions showed increases in MLD (P = .0006), DNORM (P = .001), and average diameter (P = .001). On-trial apolipoprotein (apo) B, apo C-III, and blood pressure levels inversely correlated with these compensatory changes. CONCLUSIONS Lesion progression in one coronary segment is associated with significant increases in segmental diameter of remote parts of the coronary tree. We hypothesize these increases to be vascular compensatory changes in response to progression of CAD. Vascular compensatory change is enhanced by LDL cholesterol and triglyceride-rich lipoprotein reduction and appears to be part of the treatment effect itself.

Synthesis of 3D arterial images from multiple ultrasound images

A method is outlined for generating 3-D ultrasound images of arteries from multiple 2-D images with known spatial coordinates. A magnetic spatial locating device is attached to the ultrasound transducer to track its position. Images are acquired that cover the artery from varying views. The images are digitized and then processed by computer methods to find the arterial wall boundaries. Detected boundaries from sequences of images are then transformed to a common spatial reference point and combined to represent the wall in three dimensions. The method is tested on both a phantom and on human carotid artery.<<ETX>>

Computer quantitative measurements of CLAS coronary angiograms compared to evaluation by panels of human angiographers

CLAS, a placebo controlled trial of Colestipol/Niacin therapy in men with previous coronary artery bypass surgery, obtained 188 baseline, 162 two year and 103 four year angiograms. Human panel readers evaluated all films with treatment assignment and film order masked. All lesions in native arteries and grafts were identified, assigned percent stenosis estimates, lesion change estimates and an overall change score. In a 50% random sample, all processable segments have been computer evaluated with multiple lesion and segment based algorithms derived by the USC-Caltech/JPL method. Among nonocclusive lesions, 350 were evaluated for percent stenosis by panel and computer methods with correlation between the two methods of 0.70 p < 0.0001). Four hundred forty-two lesions (principally more severe) could only be evaluated by humans. Seventy-nine early lesions could be evaluated only by computer. Seven hundred thirty-one native coronary segments were analyzed by QCA including 302 segments without discrete lesions in addition to the 429 segments with discrete lesions that were compared with the panel evaluation. After accounting for baseline effects, position relative to grafts and within subject correlation, significant therapy effects were demonstrated by computer measures of both stenosis and vessel edge roughness. These data confirm that computer analsis of angiograms can reduce subject numbers compared to human panel reading and offer different endpoint measures.

3-D Display Of Magnetic Resonance Images

As an aid for visualization, planar magnetic resonance images from human legs are used to assemble 3-D presentations of individual leg muscles. This technique is applied to two areas: (1) the study of muscle atrophy in microgravity environment and (2) the measurement of physiological cross-sectional area to evaluate muscle strength.