Quantitative measurements in IVUS images

Comparison of deep learning-based image segmentation methods for intravascular ultrasound on retrospective and large image cohort study

OBJECTIVES

The aim of this study was to investigate the generalization performance of deep learning segmentation models on a large cohort intravascular ultrasound (IVUS) image dataset over the lumen and external elastic membrane (EEM), and to assess the consistency and accuracy of automated IVUS quantitative measurement parameters.

METHODS

A total of 11,070 IVUS images from 113 patients and pullbacks were collected and annotated by cardiologists to train and test deep learning segmentation models. A comparison of five state of the art medical image segmentation models was performed by evaluating the segmentation of the lumen and EEM. Dice similarity coefficient (DSC), intersection over union (IoU) and Hausdorff distance (HD) were calculated for the overall and for subsets of different IVUS image categories. Further, the agreement between the IVUS quantitative measurement parameters calculated by automatic segmentation and those calculated by manual segmentation was evaluated. Finally, the segmentation performance of our model was also compared with previous studies.

RESULTS

CENet achieved the best performance in DSC (0.958 for lumen, 0.921 for EEM) and IoU (0.975 for lumen, 0.951 for EEM) among all models, while Res-UNet was the best performer in HD (0.219 for lumen, 0.178 for EEM). The mean intraclass correlation coefficient (ICC) and Bland-Altman plot demonstrated the extremely strong agreement (0.855, 95% CI 0.822-0.887) between model's automatic prediction and manual measurements.

CONCLUSIONS

Deep learning models based on large cohort image datasets were capable of achieving state of the art (SOTA) results in lumen and EEM segmentation. It can be used for IVUS clinical evaluation and achieve excellent agreement with clinicians on quantitative parameter measurements.

Plaque burden estimated from optical coherence tomography with deep learning: In vivo validation using co-registered intravascular ultrasound

OBJECTIVES

The objective of the present study was to compare plaque burden (PB) calculated from optical coherence tomography (OCT) using deep learning (DL) with PB derived from co-registered intravascular ultrasound (IVUS).

BACKGROUND

A DL algorithm was developed for automated plaque characterization and PB quantification from OCT images. However, the performance of this algorithm for PB quantification has not been validated.

METHODS

Five-year follow-up OCT and IVUS images from 15 patients implanted with bioresorbable vascular scaffold (BVS) at baseline were analyzed. Precise co-registration for 72 anatomical slices was achieved utilizing unique BVS radiopaque markers. PB derived from OCT DL and IVUS were compared. OCT cross-sections were divided into four subgroups with different media visibility level. The impact of media visibility on the numerical difference between OCT-derived and IVUS-derived PB was investigated. The stent sizes selected by OCT DL and IVUS were compared.

RESULTS

Sixty-four paired OCT and IVUS cross-sections were compared. OCT DL showed good concordance with IVUS for PB assessment (ICC = 0.81, difference = -3.53 ± 6.17%, p < 0.001). The numerical difference between OCT DL-derived PB and IVUS-derived PB was not substantially impacted by missing segments of media visualization (p = 0.21). OCT DL showed a diagnostic accuracy of 92% in identifying PB > 65%. The stent sizes selected by OCT DL were smaller compared to the ones selected by IVUS (difference = 0.30 ± 0.34 mm, p < 0.001).

CONCLUSIONS

The DL algorithm provides a feasible and reliable method for automated PB estimation from OCT, irrespective of media visibility. OCT DL showed good diagnostic accuracy in identifying PB > 65%, revealing its potential to complement conventional OCT imaging.

The Evolution of Data Fusion Methodologies Developed to Reconstruct Coronary Artery Geometry From Intravascular Imaging and Coronary Angiography Data: A Comprehensive Review

Understanding the mechanisms that regulate atherosclerotic plaque formation and evolution is a crucial step for developing treatment strategies that will prevent plaque progression and reduce cardiovascular events. Advances in signal processing and the miniaturization of medical devices have enabled the design of multimodality intravascular imaging catheters that allow complete and detailed assessment of plaque morphology and biology. However, a significant limitation of these novel imaging catheters is that they provide two-dimensional (2D) visualization of the lumen and vessel wall and thus they cannot portray vessel geometry and 3D lesion architecture. To address this limitation computer-based methodologies and user-friendly software have been developed. These are able to off-line process and fuse intravascular imaging data with X-ray or computed tomography coronary angiography (CTCA) to reconstruct coronary artery anatomy. The aim of this review article is to summarize the evolution in the field of coronary artery modeling; we thus present the first methodologies that were developed to model vessel geometry, highlight the modifications introduced in revised methods to overcome the limitations of the first approaches and discuss the challenges that need to be addressed, so these techniques can have broad application in clinical practice and research.

Segmenting breast cancerous regions in thermal images using fuzzy active contours

Breast cancer is the main cause of death among young women in developing countries. The human body temperature carries critical medical information related to the overall body status. Abnormal rise in total and regional body temperature is a natural symptom in diagnosing many diseases. Thermal imaging (Thermography) utilizes infrared beams which are fast, non-invasive, and non-contact and the output created images by this technique are flexible and useful to monitor the temperature of the human body. In some clinical studies and biopsy tests, it is necessary for the clinician to know the extent of the cancerous area. In such cases, the thermal image is very useful. In the same line, to detect the cancerous tissue core, thermal imaging is beneficial. This paper presents a fully automated approach to detect the thermal edge and core of the cancerous area in thermography images. In order to evaluate the proposed method, 60 patients with an average age of 44/9 were chosen. These cases were suspected of breast tissue disease. These patients referred to Tehran Imam Khomeini Imaging Center. Clinical examinations such as ultrasound, biopsy, questionnaire, and eventually thermography were done precisely on these individuals. Finally, the proposed model is applied for segmenting the proved abnormal area in thermal images. The proposed model is based on a fuzzy active contour designed by fuzzy logic. The presented method can segment cancerous tissue areas from its borders in thermal images of the breast area. In order to evaluate the proposed algorithm, Hausdorff and mean distance between manual and automatic method were used. Estimation of distance was conducted to accurately separate the thermal core and edge. Hausdorff distance between the proposed and the manual method for thermal core and edge was 0.4719 ± 0.4389, 0.3171 ± 0.1056 mm respectively, and the average distance between the proposed and the manual method for core and thermal edge was 0.0845 ± 0.0619, 0.0710 ± 0.0381 mm respectively. Furthermore, the sensitivity in recognizing the thermal pattern in breast tissue masses is 85 % and its accuracy is 91.98 %.A thermal imaging system has been proposed that is able to recognize abnormal breast tissue masses. This system utilizes fuzzy active contours to extract the abnormal regions automatically.

Graph-based IVUS segmentation with efficient computer-aided refinement

A new graph-based approach for segmentation of luminal and external elastic lamina (EEL) surface of coronary vessels in gated 20 MHz intravascular ultrasound (IVUS) image sequences (volumes) is presented. The approach consists of a fully automated segmentation stage ("new automated" or NA) and a user-guided computer-aided refinement ("new refinement" or NR) stage. Both approaches are based on the LOGISMOS approach for simultaneous dual-surface graph-based segmentation. This combination allows the user to efficiently combine general information about IVUS image appearance and case-specific IVUS morphology and therefore deal with frequently occurring issues like calcified plaque-causing signal shadowing-and imaging artifacts. The automated segmentation stage starts with pre-segmenting the lumen to automatically define the lumen centerline, which is used to transform the segmentation task into a LOGISMOS-family graph optimization problem. Following the automated segmentation, the user can inspect the result and correct local or regional segmentation inaccuracies by (iteratively) providing approximate clues regarding the location of the desired surface locations. This expert information is utilized to modify the previously calculated cost functions, locally re-optimizing the underlying modified graph without a need to start the new optimization from scratch. Validation of our method was performed on 41 gated 20 MHz IVUS data sets for which an expert-defined independent standard was available. Resulting from the automated stage of the approach (NA), the mean and standard deviation of the root mean square area errors for the luminal and external elastic lamina surfaces were 1.12 ±0.67 mm (2) and 2.35 ±1.61 mm (2) , respectively. Following the refinement stage (NR), the root mean square area errors significantly decreased to 0.82 ±0.44 mm (2) and 1.17 ±0.65 mm (2) for the same surfaces, respectively ( for both surfaces). The approach is delivering a previously unachievable speed of obtaining clinically relevant segmentations compared to the current approaches of automated segmentation followed by manual editing.

Detection of coronary plaques using MR coronary vessel wall imaging: validation of findings with intravascular ultrasound

OBJECTIVES

Compared with X-ray coronary angiography (CAG), magnetic resonance imaging of the coronary vessel wall (MR-CVW) may provide more information about plaque burden and coronary remodelling. We compared MR-CVW with intravascular ultrasound (IVUS), the standard of reference for coronary vessel wall imaging, with regard to plaque detection and wall thickness measurements.

METHODS

In this study 17 patients with chest pain, who had been referred for CAG, were included. Patients underwent IVUS and MR-CVW imaging of the right coronary artery (RCA). Subsequently, the coronary vessel wall was analysed for the presence and location of coronary plaques.

RESULTS

Fifty-two matching RCA regions of interest were available for comparison. There was good agreement between IVUS and MR-CVW for qualitative assessment of presence of disease, with a sensitivity of 94% and specificity of 76%. Wall thickness measurements demonstrated a significant difference between mean wall thickness on IVUS and MR-CVW (0.48 vs 1.24 mm, P < 0.001), but great heterogeneity between wall thickness measurements, resulting in a low correlation between IVUS and MR-CVW.

CONCLUSIONS

MR-CVW has high sensitivity for the detection of coronary vessel wall thickening in the RCA compared with IVUS. However, the use of MRI for accurate absolute wall thickness measurements is not supported when a longitudinal acquisition orientation is used.

A novel three-dimensional quantitative coronary angiography system: In-vivo comparison with intravascular ultrasound for assessing arterial segment length

BACKGROUND

Accurate on-line assessments of vessel dimensions are of utmost importance for selecting the appropriate stent size in coronary interventions. Recently a new three-dimensional quantitative coronary angiography (3D QCA) analytical software package was developed to accurately assess the vessel dimensions for the planning and guidance of such coronary interventions. This study aimed to validate the 3D QCA software package for assessing arterial segment length by comparing with intravascular ultrasound (IVUS). In addition, the difference in the two measurements from 3D QCA and IVUS for curved segments was studied.

METHODS

A retrospective study including 20 patients undergoing both coronary angiography and IVUS examinations of the left coronary artery was set up for the validation. The same vessel segments of interest between proximal and distal markers were identified and measured on both angiographic and IVUS images, by the 3D QCA software and by a quantitative IVUS software package, respectively. In addition, the curvature of each of the segments of interest was assessed and the correlation between the accumulated curvature of the segment and the difference in segment lengths measured from the two imaging modalities was analyzed.

RESULTS

37 vessel segments of interest were identified from both angiographic and IVUS images. The 3D QCA segment length was slightly longer than the IVUS segment length (15.42 +/- 6.02 mm vs. 15.12 +/- 5.81 mm, P = 0.040). The linear correlation of the two measurements was: 3D QCA Length = -0.09 + 1.03 x IVUS Length (r(2) = 0.98, P < 0.001). Bland-Altman plot showed that the difference in the two measurements was not correlated with the average of the two measurements (P = 0.141), but with the accumulated curvature of the segment (P = 0.015). After refining the difference by the correlation, the average difference of the two measurements decreased from 0.30 +/- 0.86 mm (P = 0.040) to 0.00 +/- 0.78 mm (P = 0.977).

CONCLUSIONS

The 3D QCA software package can accurately assess the actual arterial segment length. The difference in segment lengths measured from 3D QCA and IVUS was correlated with the accumulated curvature of the segment.

Impact of analyzing less image frames per segment for radiofrequency-based volumetric intravascular ultrasound measurements in mild-to-moderate coronary atherosclerosis

Volumetric radiofrequency-based intravascular ultrasound (RF–IVUS) data of coronary segments are increasingly used as endpoints in serial trials of novel anti-atherosclerotic therapies. In a relatively time-consuming process, vessel and lumen contours are defined; these contours are first automatically detected, then visually checked, and finally (in most cases) manually edited to generate reliable volumetric data of vessel geometry and plaque composition. Reduction in number of cross-sectional images for volumetric analysis could save analysis time but may also increase measurement variability of volumetric data. To assess whether a 50% reduction in number of frames per segment (every second frame) alters the reproducibility of volumetric measurements, we performed repeated RF–IVUS analyses of 15 coronary segments with mild-to-moderate atherosclerosis (20.2 ± 0.2 mm-long segments with 46 ± 13% plaque burden). Volumes were calculated based on a total of 731 image frames. Reducing the number of cross-sectional image frames for volumetric measurements saved analysis time (38 ± 9 vs. 68 ± 17 min/segment; P < 0.0001) and resulted for only a few parameters in (borderline) significant but mild differences versus measurements based on all frames (fibrous volume, P < 0.05; necrotic-core volume, P = 0.07). Compared to the intra-observer variability, there was a mild increase in measurement variability for most geometrical and compositional volumetric RF–IVUS parameters. In RF–IVUS studies of mild-to-moderate coronary disease, analyzing less image frames saved analysis time, left most volumetric parameters greatly unaffected, and resulted in a no more than mild increase in measurement variability of volumetric data.

Statistical strategy for anisotropic adventitia modelling in IVUS

Vessel plaque assessment by analysis of intravascular ultrasound sequences is a useful tool for cardiac disease diagnosis and intervention. Manual detection of luminal (inner) and media-adventitia (external) vessel borders is the main activity of physicians in the process of lumen narrowing (plaque) quantification. Difficult definition of vessel border descriptors, as well as, shades, artifacts, and blurred signal response due to ultrasound physical properties trouble automated adventitia segmentation. In order to efficiently approach such a complex problem, we propose blending advanced anisotropic filtering operators and statistical classification techniques into a vessel border modelling strategy. Our systematic statistical analysis shows that the reported adventitia detection achieves an accuracy in the range of interobserver variability regardless of plaque nature, vessel geometry, and incomplete vessel borders.

Young's modulus reconstruction of vulnerable atherosclerotic plaque components using deformable curves

Rupture, with subsequent thrombosis, of thin-cap fibroatheromas (TCFAs) is a major cause of myocardial infarction. A TCFA has two main components: these are a large, soft lipid pool and a thin, stiff fibrous cap covering it. Quantification of their morphology and stiffness is essential for monitoring atherosclerosis and quantifying the effect of plaque-stabilizing pharmaceutical treatment. To accomplish this, we have developed a model-based Young's modulus reconstruction method. From a plaque strain elastogram, measured with an intravascular ultrasound catheter, it reconstructs a Young's modulus image of the plaque. To this end, a minimization algorithm automatically varies the morphology and stiffness parameters of a TCFA computer model, until the corresponding computer-simulated strain elastogram resembles the measured strain elastogram. The morphology parameters of the model are the control-points of two deformable Bézier curves; one curve delineates the distal border of the lipid pool region, the other the distal border of the cap region. These component regions are assumed to be homogeneous and their stiffness is characterized by a Young's modulus. Reconstructions from strain elastograms that were 1. simulated using a histology-derived computer TCFA, 2. measured from a physical phantom with a soft lipid pool, and 3. simulated with a computer TCFA, where the complexity of its plaque component borders was increased, demonstrated the superior reconstruction/delineation behavior of this method, compared with a previously developed circular reconstruction method that used only circles for border delineation. Consequently, this method may become a valuable tool for the quantification of both the morphology and stiffness of vulnerable atherosclerotic plaque components.

Vasa vasorum imaging: A new window to the clinical detection of vulnerable atherosclerotic plaques

Complications of vulnerable atherosclerotic plaques (rupture, luminal and mural thrombosis, intraplaque hemorrhage, rapid progression to stenosis, spasm, and so forth) lead to heart attacks and strokes. It remains difficult to identify what plaques are vulnerable to these complications. Despite recent developments such as thermography, spectroscopy, and magnetic resonance imaging, none of them is approved for clinical use. Intravascular ultrasound (IVUS), a relatively old yet widely available clinical tool for guiding intracoronary procedures, is increasingly used for characterization of atherosclerotic plaques. However, inability of IVUS in measuring plaque activity limits its value in detection of vulnerable plaques. In this review, we present new information suggesting that microbubble contrast-enhanced IVUS can measure activity and inflammation within atherosclerotic plaques by imaging vasa vasorum density. An increasing body of evidence indicates that vasa vasorum density may be a strong marker for plaque vulnerability. We suggest that a combination of structural assessment (cap thickness, lipid core, calcification, etc) and vasa vasorum density imaging by IVUS can serve as the most powerful clinically available tool for characterization of vulnerable plaques. Due to space limitations, all IVUS images and movies are posted on the website of the Ultimate IVUS Collaborative Project: http://www.ultimateivus.com.

Quantification of dose perturbation by plaque in vascular brachytherapy

BACKGROUND

Dose prescription and reporting in vascular brachytherapy (VBT) is based on the assumption that the vessel wall is water equivalent, which does not consider a possible dose perturbation by plaque. As the extent of this perturbation is unknown, we aimed to quantify dose attenuation by atherosclerotic plaque for beta- and gamma-radiation.

MATERIAL AND METHODS

The dose delivered from Strontium-90/Yttrium-90 ((90)Sr/Y) and Iridium-192 ((192)Ir) sources with and without human peripheral arteries ((90)Sr/Y: n = 38, (192)Ir: n = 7) surrounding the respective delivery catheter was determined with radiochromic films. Plaque and vessel wall thickness were measured using light microscopy. From the ratio-attenuated doseunattenuated dose (dose perturbation factor: DPF) we determined averaged attenuation coefficients for atherosclerotic plaque (micro(P)) and the residual part of the vessel wall (micro(W)) by regression analysis based on the function DPF = exp(-micro(P) * plaque thickness -micro(W) * residual wall thickness).

RESULTS

Attenuation in case of (192)Ir was less than the measurement uncertainties. For beta-radiation correlation was found by discrimination between calcified and noncalcified plaque. Classifying noncalcified plaque as normal arterial tissue, the regression coefficient was r = 0.845 at micro(P)= 0.5356 mm(-1) and micro(W) = 0.0663 mm(-1).

CONCLUSIONS

Vascular brachytherapy with beta radiation in calcified arteries results in significant dose attenuation within the vessel wall, which can be calculated on knowing the vascular morphometry. Thus, plaque thickness should be taken into account in treatment planning and retrospective analyses.

Intravascular ultrasound assessment of ambiguous coronary lesions

Accurate assessment of coronary lesions is essential for clinical decision-making. While angiography has long been accepted as the gold standard investigation, this technique provides only a planar 2-D silhouette of the arterial lumen and therefore has limited accuracy in the setting of vessel tortuosity or overlap, bifurcational and eccentric lesions, and diffusely diseased arteries. By providing high-resolution cross-sectional imaging through the arterial wall, intravascular ultrasound (IVUS) can overcome many of these limitations and accurately quantify angiographically indeterminate lesions. Angiographic evaluation of the left main coronary artery presents particular challenges that are ideally resolved with IVUS examination. The role of IVUS in the assessment of coronary stenoses of angiographically intermediate severity (50-70%) continues to evolve. Recent data correlating IVUS with intracoronary flow and pressure measurements suggest that epicardial coronary artery lesions with minimum lumen area of less than 3-4 mm2 may be haemodynamically significant. In addition to accurately quantifying minimum lumen diameter and area at the lesion site, IVUS can characterise coronary artery plaque morphology, and it may have the potential to predict plaque complications.