Deep learning strategy for accurate carotid intima-media thickness measurement: An ultrasound study on Japanese diabetic cohort.

MOTIVATION: The carotid intima-media thickness (cIMT) is an important biomarker for cardiovascular diseases and stroke monitoring. This study presents an intelligence-based, novel, robust, and clinically-strong strategy that uses a combination of deep-learning (DL) and machine-learning (ML) paradigms. METHODOLOGY: A two-stage DL-based system (a class of AtheroEdge™ systems) was proposed for cIMT measurements. Stage I consisted of a convolution layer-based encoder for feature extraction and a fully convolutional network-based decoder for image segmentation. This stage generated the raw inner lumen borders and raw outer interadventitial borders. To smooth these borders, the DL system used a cascaded stage II that consisted of ML-based regression. The final outputs were the far wall lumen-intima (LI) and media-adventitia (MA) borders which were used for cIMT measurements. There were two sets of gold standards during the DL design, therefore two sets of DL systems (DL1 and DL2) were derived. RESULTS: A total of 396 B-mode ultrasound images of the right and left common carotid artery were used from 203 patients (Institutional Review Board approved, Toho University, Japan). For the test set, the cIMT error for the DL1 and DL2 systems with respect to the gold standard was 0.126 ±â€¯0.134 and 0.124 ±â€¯0.100 mm, respectively. The corresponding LI error for the DL1 and DL2 systems was 0.077 ±â€¯0.057 and 0.077 ±â€¯0.049 mm, respectively, while the corresponding MA error for DL1 and DL2 was 0.113 ±â€¯0.105 and 0.109 ±â€¯0.088 mm, respectively. The results showed up to 20% improvement in cIMT readings for the DL system compared to the sonographer's readings. Four statistical tests were conducted to evaluate reliability, stability, and statistical significance. CONCLUSION: The results showed that the performance of the DL-based approach was superior to the nonintelligence-based conventional methods that use spatial intensities alone. The DL system can be used for stroke risk assessment during routine or clinical trial modes.

Improved correlation between carotid and coronary atherosclerosis SYNTAX score using automated ultrasound carotid bulb plaque IMT measurement.

Described here is a detailed novel pilot study on whether the SYNTAX (Synergy between percutaneous coronary intervention with TAXUS and cardiac surgery) score, a measure of coronary artery disease complexity, could be better predicted with carotid intima-media thickness (cIMT) measures using automated IMT all along the common carotid and bulb plaque compared with manual IMT determined by sonographers. Three hundred seventy consecutive patients who underwent carotid ultrasound and coronary angiography were analyzed. SYNTAX score was determined from coronary angiograms by two experienced interventional cardiologists. Unlike most methods of cIMT measurement commonly used by sonographers, our method involves a computerized automated cIMT measurement all along the carotid artery that includes the bulb region and the region proximal to the bulb (under the class of AtheroEdge systems from AtheroPoint, Roseville, CA, USA). In this study, the correlation between automated cIMT that includes bulb plaque and SYNTAX score was found to be 0.467 (p < 0.0001), compared with 0.391 (p < 0.0001) for the correlation between the sonographer's IMT reading and SYNTAX score. The correlation between the automated cIMT and the sonographer's IMT was 0.882. When compared against the radiologist's manual tracings, automated cIMT system performance had a lumen-intima error of 0.007818 ± 0.0071 mm, media-adventitia error of 0.0179 ± 0.0125 mm and automated cIMT error of 0.0099 ± 0.00988 mm. The precision of automated cIMT against the manual radiologist's reading was 98.86%. This current automated algorithm revealed a significantly stronger correlation between cIMT and coronary SYNTAX score as compared with the sonographer's cIMT measurements with multiple cardiovascular risk factors. We benchmarked our correlation between the automated cIMT that includes bulb plaque and SYNTAX score against a previously published (Ikeda et al. 2013) AtheroEdgeLink (AtheroPoint) correlation between the automated cIMT that does not include bulb plaque and SYNTAX score and had an improvement of 44.58%. By sampling cIMT in the bulb region, the automated cIMT technique improves the degree of correlation between coronary artery disease lesion complexity and carotid atherosclerosis characteristics.

Shape-based approach for coronary calcium lesion volume measurement on intravascular ultrasound imaging and its association with carotid intima-media thickness.

OBJECTIVES: Coronary calcification plays an important role in diagnostic classification of lesion subsets. According to histopathologic studies, vulnerable atherosclerotic plaque contains calcified deposits, and there can be considerable variation in the extent and degree of calcification. Intravascular ultrasound (IVUS) has demonstrated its role in imaging coronary arteries, thereby displaying calcium lesions. The aim of this work was to develop a fully automated system for detection, area and volume measurement, and characterization of the largest calcium deposits in coronary arteries. Furthermore, we demonstrate the correlation between the coronary calcium IVUS volume and the neurologic risk biomarker B-mode carotid intima-media thickness (IMT). METHODS: Our system automatically detects the frames with calcium, identifies the largest calcium region, and performs shape-based volume measurements. The carotid IMT is measured by using AtheroEdge software (AtheroPoint, LLC) on B-mode ultrasound imaging. RESULTS: Our database consists of low-contrast IVUS videos and corresponding B-mode images from 100 patients. Our experiments showed that the correlation between calcium volumes and carotid IMT was higher for the left carotid artery compared to the right carotid artery (r = 0.066 for the left carotid artery and 0.121 for the right carotid artery). We obtained 97% accuracy for automated calcium detection compared against the scoring given by our expert radiologists. Furthermore, we benchmarked shape-based volume measurement against the conventional method, which used integration of regions and showed a correlation of 84%. CONCLUSIONS: Since carotid IMT is an independent prognostic factor for myocardial infarction, and calcium lesions are correlated with stroke risk, we believe that this automated system for calcium volume measurement could be useful for assessing patients' cardiovascular risk.

A comparative approach of four different image registration techniques for quantitative assessment of coronary artery calcium lesions using intravascular ultrasound.

In IVUS imaging, constant linear velocity and a constant angular velocity of 1800 rev/min causes displacement of the calcium in subsequent image frames. To overcome this error in intravascular ultrasound video, IVUS image frames must be registered prior to the lesion quantification. This paper presents a comprehensive comparison of four registration methods, namely: Rigid, Affine, B-Splines and Demons on five set of calcium lesion quantification parameters namely: (i) the mean lesion area, (ii) mean lesion arc, (iii) mean lesion span, (iv) mean lesion length, and (v) mean lesion distance from catheter. Using our IRB approved data of 100 patient volumes, our results shows that all four registrations showed a decrease in five calcium lesion parameters as follows: for Rigid registration, the values were: 4.92%, 5.84%, 5.89%, 5.27%, and 4.57%, respectively, for Affine registration the values were: 6.06%, 6.51%, 7.28%, 6.50%, and 5.94%, respectively, for B-Splines registration the values were: 7.35%, 8.03%, 9.54%, 8.18%, and 7.62%, respectively, and for Demons registration the five parameters were 7.32%, 8.02%, 10.11%, 7.94%, and 8.92% respectively. The relative overlap of identified lesions decreased by 5.91% in case of Rigid registration, 6.23% in case of Affine registration, 4.48% for Demons registration, whereas it increased by 3.05% in case of B-Splines registration. Rigid and Affine transformation-based registration took only 0.1936 and 0.2893 s per frame, respectively. Demons and B-Splines framework took only 0.5705 and 0.9405 s per frame, respectively, which were significantly slower than Rigid and Affine transformation based image registration.

Carotid artery distensibility: a reliability study.

OBJECTIVE: Carotid distensibility (CD) is a measure of carotid artery elasticity that has been introduced as a risk factor for cardiovascular disease. Information regarding reproducibility of sonographic CD measures is limited. The objective of this study was to evaluate the inter-reader reliability of sonographic measurements of common carotid artery (CCA) diameters and derived metrics of CD. METHODS: Two independent readers (R1 and R2) measured the systolic diameter (SD) and diastolic diameter (DD) for the right CCA from the B/M-mode sonographic registrations among 118 subjects. The derived CD metrics (strain, elastic modulus [E], stiffness [beta], and CD) were calculated. The inter-reader type 3 intraclass correlation coefficients (ICC3,1) for carotid diameters were calculated. RESULTS: The mean SDs +/- standard deviation were 7.15 +/- 1.43 mm for R1 and 7.24 +/- 1.43 mm for R2. The mean DDs were 6.71 +/- 1.36 mm for R1 and 6.68 +/- 1.41 mm for R2. The mean differences of SD and DD between R1 and R2 were 0.08 +/- 0.40 mm (paired t test, P = .04) and 0.03 +/- 0.43 mm (paired t test, P = .46), respectively. Inter-reader type 3 intraclass correlation coefficients were 0.96 for SD and 0.95 for DD. We observed a significant association of demographics with carotid diameters but not with derived CD metrics or risk factors. CONCLUSIONS: Our results suggest good reproducibility of CCA diameters measured with B/M-mode sonography. However, very small changes in linear measurements of carotid diameters can have big effects on estimates of arterial mechanical properties such as strain and Young's modulus. The standard boundary identification methods may not be precise and reproducible enough for use in a clinical setting.