Anuja Nair

Coronary Plaque Classification With Intravascular Ultrasound Radiofrequency Data Analysis

Background—Atherosclerotic plaque stability is related to histological composition. However, current diagnostic tools do not allow adequate in vivo identification and characterization of plaques. Spectral analysis of backscattered intravascular ultrasound (IVUS) data has potential for real-time in vivo plaque classification. Methods and Results—Eighty-eight plaques from 51 left anterior descending coronary arteries were imaged ex vivo at physiological pressure with the use of 30-MHz IVUS transducers. After IVUS imaging, the arteries were pressure-fixed and corresponding histology was collected in matched images. Regions of interest, selected from histology, were 101 fibrous, 56 fibrolipidic, 50 calcified, and 70 calcified-necrotic regions. Classification schemes for model building were computed for autoregressive and classic Fourier spectra by using 75% of the data. The remaining data were used for validation. Autoregressive classification schemes performed better than those from classic Fourier spectra with accuracies of 90.4% for fibrous, 92.8% for fibrolipidic, 90.9% for calcified, and 89.5% for calcified-necrotic regions in the training data set and 79.7%, 81.2%, 92.8%, and 85.5% in the test data, respectively. Tissue maps were reconstructed with the use of accurate predictions of plaque composition from the autoregressive classification scheme. Conclusions—Coronary plaque composition can be predicted through the use of IVUS radiofrequency data analysis. Autoregressive classification schemes performed better than classic Fourier methods. These techniques allow real-time analysis of IVUS data, enabling in vivo plaque characterization.

Assessing spectral algorithms to predict atherosclerotic plaque composition with normalized and raw intravascular ultrasound data.

Spectral analysis of backscattered intravascular ultrasound (IVUS) data has demonstrated the ability to characterize plaque. We compared the ability of spectral parameters (e.g., slope, midband fit and y-intercept), computed via classic Fourier transform (CPSD), Welch power spectrum (WPSD) and autoregressive (MPSD) models, to classify plaque composition. Data were collected ex vivo from 32 human left anterior descending coronary arteries. Regions-of-interest (ROIs), selected from histology, comprised 64 collagen-rich, 24 fibrolipidic, 23 calcified and 37 calcified-necrotic regions. A novel quantitative method was used to correlate IVUS data with corresponding histologic sections. Periodograms of IVUS samples, identified for each ROI, were used to calculate spectral parameters. Statistical classification trees (CT) were computed with 75% of the data for plaque characterization. The remaining data were used to assess the accuracy of the CTs. The overall accuracies for normalized spectra with CPSD, WPSD and MPSD were, respectively, 84.7%, 85.6% and 81.1% (training data) and 54.1%, 64.9% and 37.8% (test data). These numbers were improved to 89.2%, 91.9% and 89.2% (training) and 62.2%, 73% and 59.5% (test) when the calcified and calcified-necrotic regions were combined for analysis. Most CTs misclassified a few fibrolipidic regions as collagen, which is histologically acceptable, and the unnormalized and normalized spectra results were similar.

Regularized autoregressive analysis of intravascular ultrasound backscatter: improvement in spatial accuracy of tissue maps

Autoregressive (AR) models are qualified for analysis of stochastic, short-time data, such as intravascular ultrasound (IVUS) backscatter. Regularization is required for AR analysis of short data lengths with an aim to increase spatial accuracy of predicted plaque composition and was achieved by determining suitable AR orders for short data records. Conventional methods of determining order were compared to the use of trend in the mean square error for determining order. Radio-frequency data from 101 fibrous, 56 fibro-lipidic, 50 calcified, and 70 lipid-core regions of interest (ROIs) were collected ex vivo from 51 human coronary arteries with 30 MHz unfocused IVUS transducers. Spectra were computed for AR model orders between 3-20 for data representing ROIs of two sizes (32 and 16 samples at 100 MHz sampling frequency) and were analyzed in the 17-42 MHz bandwidth. These spectra were characterized based on eight previously identified parameters. Statistical classification schemes were computed from 75% of the data and cross-validated with the remaining 25% using matched histology. The results determined the suitable AR order numbers for the two ROI sizes. Conventional methods of determining order did not perform well. Trend in the mean square error was identified as the most suitable factor for regularization of short record lengths.

Potential of dual-energy computed tomography to characterize atherosclerotic plaque: ex vivo assessment of human coronary arteries in comparison to histology

BACKGROUND Noninvasive characterization of coronary atherosclerotic plaque is limited with current computed tomography (CT) techniques. Dual-energy CT (DECT) has the potential to provide additional attenuation data for better differentiation of plaque components. OBJECTIVE We attempted to characterize coronary atherosclerotic plaque with DECT. METHODS Seven human coronary arteries acquired at autopsy were scanned consecutively at 80 and 140 kVp with CT. Vessels were perfused with saline, and data were acquired before and after contrast agent injection. Lesions were identified, and attenuation measurements were made from CT image quadrants. CT quadrants were classified as densely calcified, fibrocalcific, fibrous, lipid-rich, or normal vessel wall, corresponding to matched histology images. Attenuation values at each peak tube voltage were compared within plaque types for both noncontrast and contrast scans. Further, dual-energy index (DEI) values computed from attenuation were analyzed for classification of plaque. RESULTS In 14 lesions, a total of 56 quadrants were identified. Histology results classified 8 (14%) as densely calcified, 8 (14%) as fibrocalcific, 9 (16%) as fibrous, 5 (9%) as lipid-rich, and 25 (45%) as normal vessel wall. Calcified lesions attenuated significantly more at 80 kVp in both contrast and noncontrast scans, whereas fibrous plaque attenuated more at 80 kVp only for contrast-enhanced scans. No differences were found for lipid-rich plaques. Using DEI values, only densely calcified plaques could be distinguished from other plaque types except fibrocalcific plaques in contrast images. CONCLUSIONS Only densely calcified and fibrocalcific plaques showed a true change in attenuation at 80 versus 140 kVp. Therefore, calcified plaques could be distinguished from noncalcified plaques with DECT, but further classification of plaque types was not possible.

Contrast enhancement of coronary atherosclerotic plaque: a high-resolution, multidetector-row computed tomography study of pressure-perfused, human ex-vivo coronary arteries.

ObjectivesThe objective of this study was to investigate the effect of contrast injection on atherosclerotic coronary plaque attenuation measured using multidetector-row computed tomography. BackgroundRecent multidetector-row computed tomography studies have described the characterization of coronary atherosclerotic plaque on the basis of Hounsfield unit values. The influence of contrast injection on the attenuation of individual plaque components, however, is unknown. MethodsUsing a pressurized perfusion system, 10 human coronary arteries were examined postmortem with multidetector-row computed tomography and histology. Pre-enhanced, peak-enhanced, and delayed enhanced multidetector-row computed tomography images were acquired during continuous perfusion of the vessel. A total of 37 focal atherosclerotic plaques were identified. Vessel wall attenuation was measured from multidetector-row computed tomography images during all three enhancement phases. On the basis of the histology, plaques were categorized as noncalcified (predominantly fibrous or predominantly fibrofatty), mixed calcified (calcified fibrous or calcified necrotic core), or densely calcified. The mean Hounsfield unit was compared among contrast phases for all plaques and in plaque subgroups. ResultsWe observed contrast enhancement of atherosclerotic plaques within the vessel wall. For noncalcified plaques including both fibrous and fibrofatty plaques, the mean Hounsfield unit of the vessel wall during and after contrast injection exceeded the mean value before injection (t-test, P<0.002). ConclusionThe present study demonstrates that intra-arterial injection of iodinated contrast agent results not only in luminal enhancement but also in atherosclerotic plaque enhancement in pressure-perfused coronary arteries imaged ex vivo. Plaque enhancement should be considered when characterizing plaque components on the basis of Hounsfield unit with multidetector-row computed tomography.

Intravascular ultrasound measures of coronary atherosclerosis are associated with the Framingham risk score: an analysis from a global IVUS registry

AIMS In addition to an adjunctive imaging platform during coronary angiography, intravascular ultrasound (IVUS) with Virtual Histology (VH) is increasingly being used to quantify coronary atherosclerosis. The relationship between VH-IVUS measures of coronary atherosclerosis and traditional cardiovascular risk factors has not been completely described. The objective of this study was to determine if an association exists between VH-IVUS measures of coronary atherosclerosis and the Framingham risk score in a prospective, multinational registry. METHODS AND RESULTS Patients enrolled from 2004-2006 at 37 multinational centres in the prospective VHIVUS Global Registry were analysed. All subjects underwent diagnostic coronary angiography followed by IVUS. A Framingham risk score (FRS) was calculated for each subject, then stratified into three exclusive estimates (<10%, 10-19%, or >or= 20%) for future coronary heart disease (CHD) event risk over 10 years. Among 531 patients, plaque volume of the most diseased 10 mm segment increased with increasing FRS (P=0.006, adjusted for multiple comparisons). Patients with higher FRS estimates of CHD risk had a higher proportion of plaque classified as thin cap fibroatheroma compared with patients in the middle and lower risk score categories (21.4% vs 15.2% and 11.3%, respectively, P=0.008, adjusted for multiple comparisons). CONCLUSIONS Using data from a large, multinational VH-IVUS registry we describe an association between the Framingham risk score and VH-IVUS measures of atherosclerosis within the most diseased 10 mm segment, namely plaque volume and the proportion of thin cap fibroatheroma.

Segmentation of three-dimensional intravascular ultrasound images using spectral analysis and a dual active surface model

Intravascular ultrasound (IVUS) imaging provides detailed assessment of the coronary anatomy and can be used as at quantitative tool for tracking the progression of atherosclerotic disease. The geometric information also provides evaluation of positive remodeling, a phenomenon linked to the likelihood of plaque rupture. To provide this valuable information, luminal and medial-adventitial borders must be identified in the sequence of IVUS images, a problem traditionally approached using gray-scale intensity based algorithms. However, by acquiring the radiofrequency (RF) IVUS data, the frequency information that is typically ignored can be used to improve segmentation algorithms.

Choosing the optimal mother wavelet for decomposition of radio-frequency intravascular ultrasound data for characterization of atherosclerotic plaque lesions

Aim: The objective of this work is to determine the optimal basis function to perform wavelet analysis for tissue characterization of radio frequency intravascular ultrasound (IVUS) backscattered data. This is the most important step in wavelet analysis as it ensures accurate decomposition of the original signal into the various frequency bands. The criterion to choose the mother wavelet that is best suited to the data depends on the intended application. Wavelet families possessing properties like orthogonality, regularity, stability and admissibility have previously been shown to have application in tissue characterization. Algorithm: Depending on the usable data bandwidth known from previous studies we decomposed data using a 4-level decomposition scheme. We then calculated Shannon’s entropy for every level and employed “minimum Shannon entropy criterion” to determine the best mother wavelet for signal decomposition. According to this criterion, accurate decomposition is indicated when the total entropy of the daughter (decomposed) levels is lower than the entropy of the parent level. Analysis and Results: We acquired 40 MHz IVUS data ex-vivo from 10 left anterior descending (LAD) coronary arteries. Data was acquired such that each frame comprised of 256 scanlines. Next, we randomly selected 3 scanlines for each LAD and applied the above-mentioned Shannon entropy criterion for these 30 scanlines. We analyzed 23 mother wavelets from different families. Daubechies 3rd order wavelet accurately decomposes 29/30 scanlines at all levels. Daubechies 6th order wavelet appears optimal for 21/30 scanlines. Future direction: To obtain more precise signal decomposition, the optimal mother wavelet should be selected at every decomposition level. The best mother wavelet is indicated by the lowest Shannon entropy for that particular level.

Ex vivo validation of 45 MHz intravascular ultrasound backscatter tissue characterization

AIMS The objectives of the present study are to describe the algorithm for VH(®) IVUS using the 45-MHz rotational IVUS catheter and the associated ex vivo validation in comparison to the gold standard histology. METHODS AND RESULTS The first phase of the present study was to construct the 45 MHz VH IVUS algorithm by using a total of 55 human coronary artery specimens [111 independent coronary lesions and 510 homogenous regions of interest (ROIs)], obtained at autopsy. Regions were selected from histology and matched with their corresponding IVUS data to build the plaque classification system using spectral analysis and statistical random forests. In the second phase, the ex vivo validation of the VH IVUS algorithm assessed a total of 1060 ROIs (120 lesions from 60 coronary arteries) in comparison with histology. In an independent manner, two interventional cardiologists also classified a randomly selected subset of the ROIs for assessment of inter- and intra-observer reproducibility of VH IVUS image interpretation.When including all ROIs, the predictive accuracies were 90.8% for fibrous tissue, 85.8% for fibro fatty tissue, 88.3% for necrotic core, and 88.0% for dense calcium. The exclusion of ROIs in the acoustically attenuated areas improved the predictive accuracies, ranging from 91.9 to 96.8%. The independent analysis of randomly selected 253 ROIs showed substantial agreement for inter-observer (k = 0.66) and intra-observer (k = 0.88) reproducibility. CONCLUSION Tissue classification by 45 MHz VH IVUS technology, when not influenced by calcium-induced acoustic attenuation, provided combined tissue accuracy >88% to identify tissue types compared with the gold standard histologic assessment, with high inter- and intra-observer reproducibility.

Normalization and backscatter spectral analysis of human carotid arterial data acquired using a clinical linear array ultrasound imaging system

The risk of plaque rupture in carotid atherosclerotic disease is associated more closely with the composition of plaque rather than the severity of stenosis. The constituents of plaque can be determined from ultrasonic spectral parameters obtained from normalized backscatter tissue data. Calibration of the data is done using echoes off a specular reflector which removes the system response of an ultrasound transducer, Terason™ (Teratech Corporation), from the backscatter data. A reference spectrum study is used to compare specular reflectors based on time domain (echo) and frequency domain (power spectrum, centroid and parabola test) analysis. Nylon and a tissue-mimicking phantom (velocity = 1560 m/s, slope of attenuation = 0.7 dB/cm MHz) have an intermediate acoustic impedance with respect to water and appear good choices as specular reflectors for clinical ultrasound imaging scanners compared to Plexiglas and other higher reflecting materials. A tissue-mimicking phantom is used to correct for attenuation in plaque, diffraction and saturation of electronics of the ultrasound scanner. Autoregressive power spectrum estimation methods are used to extract spectral parameters (spectral slope, y-intercept, midband fit, maximum and minimum power with corresponding frequencies, and integrated backscatter) from calibrated tissue data and linear and quadratic discriminant rules developed for classification of carotid arterial plaque. Regions of interest (n = 64; 64 samples x 8 scan lines with 30 MHz sampling frequency) consisting of 48 fibrous-fibrofatty (Class 1), 11 thrombus-necrotic core (Class 2), and 5 dense calcium (Class 3) areas selected for analysis show that fibrosis can be differentiated from necrosis and calcification. The quadratic discriminant rule identified necrosis with a lower misclassification rate (9.1%) than the linear discriminant rule (18.2%).