Abstract 14690: Deep Learning Assisted Target Lesion Analysis Using Advanced Border Detection Algorithm of Intravascular Ultrasound

Accurate segmentation of coronary arteries is important for the diagnosis of cardiovascular diseases. In this paper, we propose a fully convolutional neural network to efficiently delineate the boundaries of the wall and lumen of the coronary arteries using intravascular ultrasound (IVUS) images. Our network addresses multi-label segmentation of the wall and lumen areas at the same time. The primary body of the proposed network is U-shaped which contains the encoding and decoding paths to learn rich hierarchical representations. The multi-scale input layer is adapted to take a multi-scale input. We deploy a multi-label loss function with weighted pixel-wise cross-entropy to alleviate imbalance of the rate of background, wall, and lumen. The proposed method is compared with three existing methods and the segmentation results are measured on four metrics, dice similarity coefficient, Jaccard index, percentage of area difference, and Hausdorff distance on totally 38,478 IVUS images from 35 subjects.

Natural history of intravascular ultrasound–detected edge dissections from coronary stent deployment

Coronary plaque progression and instability are associated with expansive remodeling of the arterial wall. However, the remodeling response during plaque-stabilizing therapy and its relationship to markers of lipid metabolism and inflammation are incompletely understood. Serial intravascular ultrasound (IVUS) data from the Reversal of Atherosclerosis with Aggressive Lipid Lowering Therapy (REVERSAL) trial were obtained during 18 months of intensive versus moderate lipid-lowering therapy. In a subgroup of 210 patients, focal coronary lesions with mild luminal narrowing were identified. Lumen area, external elastic membrane (EEM) area, and plaque area were determined at the lesion and proximal reference sites at baseline and during follow-up. The remodeling ratio (RR) was calculated by dividing the lesion EEM area by the reference EEM area. The relationship between the change in remodeling, change in plaque area, lipid profile, and inflammatory markers was examined. At the lesion site, a progression in plaque area (8.9+/-25.7%) and a decrease in the RR (-3.0+/-11.2%) occurred during follow-up. In multivariable analyses, the percentage change in plaque area (P<0.0001), baseline RR (P<0.0001), baseline lesion lumen area (0.019), logarithmic value of the change in high-sensitivity C-reactive protein (P=0.027), and hypertension at baseline (P=0.014) showed a significant, direct relation with the RR at follow-up. Lesion location in the right coronary artery (P=0.006), percentage change in triglyceride levels (P=0.049), and age (P=0.037) demonstrated a significant, inverse relation with the RR at follow-up. Changes in LDL cholesterol, HDL cholesterol, and treatment group demonstrated no significant associations. Constrictive remodeling of the arterial wall was observed during plaque-stabilizing therapy with statin medications and appears related to their antiinflammatory effects.

Robust segmentation of arterial walls in intravascular ultrasound images using Dual Path U-Net

Over 35% of all US adults are estimated to have metabolic syndrome (MetS), which is defined as the presence of three or more of the following risk factors: obesity, hypertension, glucose intolerance, insulin resistance, and dyslipidemia. Furthermore, the incidence of type 2 diabetes is rapidly growing, often in an overlapping population. MetS and diabetes both substantially increase the risk of coronary artery disease (CAD), which continues to be the leading cause of death in the US. We investigated the extent to which diabetic hyperglycemia superimposed on MetS exacerbates CAD. We performed this study in the well‐characterized Ossabaw miniature swine model of MetS and CAD. The beta‐cell toxin alloxan was administered to swine by intravenous injection, then swine were placed in two groups: pigs that responded only transiently to alloxan thereafter having fasting normoglycemia (MetS, n=8) and those that responded to alloxan becoming permanently diabetic (MetS/D, n=8). After alloxan administration, swine were fed an atherogenic diet for six months to induce the MetS phenotype. At sacrifice, we used intravascular ultrasound (IVUS) imaging to quantify the extent of CAD by assessment of percent plaque burden and spotty calcification in the left anterior descending artery. Percent plaque burden quantifies the percentage of the area within the external elastic membrane (EEM) that is occupied by atherosclerotic plaque, using the formula (EEM area – lumen area)/EEM area*100%. The percent plaque burden was significantly greater in left anterior descending artery of the MetS/D swine compared to MetS swine (12.9% vs. 3.7%; p&lt;0.05). Furthermore, this difference was more prominent in the proximal 15 mm of the artery (17.1% vs. 3.9%; p&lt;0.05). We assessed coronary artery calcification by identifying echogenic signal stronger than adventitia and with acoustic shadowing peripheral to the calcium deposit. We calculated a calcium index for each artery using the formula (Frames with calcification/Total frames analyzed)*(Maximal arc angle of calcium/360°). The calcium index was significantly greater in the proximal 15 mm of the left anterior descending artery of MetS/D swine compared to MetS swine (0.0204 vs. 0.00003; p&lt;0.05). These results show significant positive correlation with Von Kossa stained sections from the proximal 15 mm of each vessel and analyzed for calcified area (r=0.84). Thus, we conclude that CAD, as measured by plaque burden and presence of calcification by IVUS, is exacerbated when diabetes is superimposed on MetS.Support or Funding InformationThis work was supported by NIH T35 HL110854, P30 DK097512, the Fortune‐Fry Ultrasound Research Fund, and Eli Lilly &amp; Co.

Background/AimsThe aim of this study was to assess the effects of a usual dose of simvastatin (20 mg/day) on plaque regression and vascular remodeling at the peri-stent reference segments after bare-metal stent implantation.MethodsWe retrospectively investigated serial intravascular ultrasound (IVUS) findings in 380 peri-stent reference segments (184 proximal and 196 distal to the stent) in 196 patients (simvastatin group, n = 132 vs. non-statin group, n = 64). Quantitative volumetric IVUS analysis was performed in 5-mm vessel segments proximal and distal to the stent.ResultsIVUS follow-up was performed at a mean of 9.4 months after stenting (range, 5 to 19 months). No significant differences were observed in the changes in mean plaque plus media (P&M) area, mean lumen area, and mean external elastic membrane (EEM) area from post-stenting to follow-up at both proximal and distal edges between the simvastatin and non-statin group. Although lumen loss within the first 3 mm from each stent edge was primarily due to an increase in P&M area rather than a change in EEM area, and lumen loss beyond 3 mm from each stent edge was due to a combination of increased P&M area and decreased EEM area, no significant differences in changes were observed in P&M, EEM, and lumen area at every 1-mm subsegment between the simvastatin and non-statin group.ConclusionsA usual dose of simvastatin does not inhibit plaque progression and lumen loss and does not affect vascular remodeling in peri-stent reference segments in patients undergoing bare-metal stent implantation.

One of the most widespread conditions that also have a high mortality rate is coronary artery disease (CAD). The accurate identification of aortic stenosis can lead to more effective therapy, which in turn can reduce the risk of dying from the condition. Images obtained using intravascular ultrasonography (IVUS) can be analyzed to determine the status of plaque that is located between the lumen and the media adventitia within the artery. To accomplish this task, IVUS's components need to be identified. In this paper, the approach to segment IVUS's component consisting of lumen, media, and adventitia is proposed. Support Vector Machine classification model is trained using extracted spatial features. Gray level cooccurrence matrix and Gabor features are extracted from polar space transformed images. The result of the proposed approach reaches 0.82 of the Jaccard Index (JM), 10.87 of Hausdorff Distance (HD), and 1.89 of Percentage Area Difference (PAD). Moreover, the classification model is evaluated using a confusion matrix. The performance of the classification is 0.89 accuracy, 0.921 precision, and 0.944 recall.

Objective To explore the characteristics of attenuated plaque (AP) in culprit lesions for acute coronary syndrome (ACS) in elderly patients. Methods This study included 166 ACS patients meeting the conditions from Jan. 1, 2013 to Sep. 31, 2014. Clinical data, vascular lesions determined by coronary angiography and intravenous ultrasound (IVUS)-demonstrated characteristics of culprit plaque, which included presence of AP, maximum attenuation arc, extravascular elastic membrane area, lumen area, plaque area, plaque burden and remodeling index, were recorded. Results Compared with non-elderly group, elderly group had a higher proportion of multiple coronary-artery lesions (50.0% vs. 35.1%, χ2=6.525, P=0.038), more attenuation plaques (62.0% vs. 45.9%, χ2=4.245, P=0.039), a larger maximum attenuation arc〔(142±80)° vs. (115±54)°, t=5.254, P=0.000〕, larger plaque area〔(14.2±3.9)mm2vs. (12.3±4.1)mm2,t=2.325, P=0.022〕, more plaque burden 〔(81.1±14.2)% vs. (76.4±13.5)%, t=2.025, P=0.042〕 and higher remodeling index 〔(1.19±0.17) vs. (1.09±0.13), t=4.245, P=0.031〕. Conclusions The onset of ACS is more closely related with the unstable plaque in elderly patients, and strengthening the measures for plaque stabilization will be helpful in the prevention and treatment of ACS in the elderly. Key words: Acute coronary syndrome; Coronary angiography; Ultrasonography; Attenuated plaque

The aim of this study was to assess whether intravascular ultrasound (IVUS)-guided angioplasty in femoropopliteal lesions would improve clinical outcomes and predict restenosis. Studies in which IVUS-guided angioplasty was used for femoropopliteal lesions were searched from the MEDLINE, Embase, Web of Science, and Cochrane databases; articles with the full text were included. The primary endpoint of this study was primary patency at 12 months, while the secondary endpoints were primary patency at 24 months, freedom from target lesion revascularization (TLR) at 12 months, and correlation of restenosis with the distal external elastic membrane (EEM) area, postintervention minimum lumen area, lesion length, dissection, and calcification. Altogether, 11 observational studies involving 1521 patients (1703 lesions) were analyzed. The quality of the evidence for 7 main outcomes was assessed as "very low" by The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) classification. The mean follow-up period was 1.5 years. The pooled rates were: 78% for 12-month primary patency (95% confidence interval [CI], 0.72-0.83), 74.3% for 24-month primary patency (95% CI: 0.71-0.78), and 80% for 12-month freedom from TLR (95% CI: 0.74-0.86). The 12-month primary patency of IVUS use (relative risk [RR], 2.01; 95% CI: 1.48-2.74) was higher compared to non-IVUS use. The minimum lumen (stent) area (standard mean difference [SMD] = -0.30; 95% CI: -0.46 to -0.15) and dissection (OR 1.58; 95% CI: 1.01-2.49, P=0.047), were associated with midterm patency in terms of restenosis. In IVUS-guided angioplasty in patients with femoropopliteal lesions, the minimum lumen (stent) area and dissection were associated with restenosis. Nevertheless, there is limited and heterogeneous evidence regarding the usefulness and Predictability of IVUS in patients with peripheral arterial disease in the femoropopliteal artery, especially in long-term patency and as a predictor of declining patency. The optimal role of IVUS in such patients should be elucidated in the future.

Deep learning-based intravascular ultrasound segmentation for the assessment of coronary artery disease

A paracrine effect of epicardial fat in underlying coronary atherosclerosis has been proposed, due to the inflammatory role of adipose tissue. Most studies evaluate globally fat (by magnetic resonance) and atherosclerosis (by computerized tomography). We analyzed microscopically, segment by segment, the potential local influence of fat in atherosclerotic plaque size and burden. Thickness of epicardial fat, coronary lumen area, and area delimited by the internal elastic lamina were measured with an image analysis system coupled to a light microscope, and plaque size (area delimited by the internal elastic lamina less lumen area) and burden (plaque area/ area delimited by the internal elastic lamina) were calculated in 192 coronary artery segments from 9 hearts from necropsies of patients that died with coronary atherosclerosis. To compare epicardial and myocardial faces of the arteries, the measures were made in each half; total amounts were also calculated. The relationships between epicardial fat and underlying atherosclerosis, both considering epicardial half and the whole segment, and between total fat and the total arterial segment, were tested by Pearson’s test. Plaque size and burden in each half were compared by paired Wilcoxon signed rank test. We considered correlation strength as r =0.10 to 0.29- weak, r = 0.30 to 0.49- medium, and r = 0.50 to 1.0- strong correlation, and results significant if P &lt; or = 0.05. Results: Correlation coefficients: epicardial face fat thickness - 1) with plaque area at epicardial half - r = 0.22, P &lt;0.01; 2) with plaque burden at epicardial half- r = 0.18, P =0.02; 3) with total plaque area- r = 0.19, P &lt;0.01; 4) with total plaque burden- r = 0.10, P =0.18; total epicardial fat thickness - 1) with total plaque area - r = 0.22, p&lt;0.01; 2) with total plaque burden- r = 0.06, P =0.39. Comparison between halves: area size - mean 1.20 mm at epicardial face and 1.5 mm at myocardial face, P =0.96; plaque burden - mean 33.7% at epicardial face and 45.7% at myocardial face, P =0.98. Conclusion: the correlations were weak and/or non-significant, and the atherosclerotic process involves equally both halves of the coronary arteries; thus our data indicate that there is no significant effect of fat tissue in underlying coronary atherosclerosis.

BackgroundThis study aimed to evaluate the relationship between cigarette smoking and coronary atherosclerotic burden, volume and composition as determined in-vivo by grayscale and virtual histology (VH) intravascular ultrasound (IVUS).Methods and ResultsBetween 2008 and 2011, (VH-)IVUS of a non-culprit coronary artery was performed in 581 patients undergoing coronary angiography. To account for differences in baseline characteristics, current smokers were matched to never smokers by age, gender and indication for catheterization, resulting in 280 patients available for further analysis. Coronary atherosclerotic plaque volume, burden, composition (fibrous, fibro-fatty, dense calcium and necrotic core) and high-risk lesions (VH-IVUS derived thin-cap fibroatheroma (TCFA), plaque burden ≥70%, minimal luminal area ≤4.0 mm2) were assessed. Cigarette smoking showed a tendency towards higher coronary plaque burden (mean±SD, 38.6±12.5% in current versus 36.4±11.0% in never smokers, p = 0.080; and odds ratio (OR) of current smoking for plaque burden above versus below the median 1.69 (1.04–2.75), p = 0.033). This effect was driven by an association in patients presenting with an acute coronary syndrome (ACS) (current smokers, plaque burden 38.3±12.8% versus never smokers, plaque burden 35.0±11.2%, p = 0.049; OR 1.88 (1.02–3.44), p = 0.042). Fibrous tissue tended to be lower in current smokers (mean±SD, 57.7±10.5% versus 60.4±12.6%, p = 0.050) and fibro-fatty tissue was higher in current smokers (median[IQR], 9.6[6.0–13.7]% versus 8.6[5.8–12.2]%, p = 0.039). However, differences in percentage necrotic core and dense calcium could not be demonstrated. Also, no differences were found with regard to high-risk lesions.ConclusionsAn association between smoking and degree of coronary atherosclerosis was present in patients undergoing coronary angiography who presented with ACS. Although smoking was associated with higher fibro-fatty percentage, no associations could be demonstrated with percentage necrotic core, nor with VH-IVUS derived TCFA lesions. Since the magnitude of the differences in both degree and composition of atherosclerosis was modest, clinical relevance of the findings may be questioned.

Predictors of Plaque Rupture Within Nonculprit Fibroatheromas in Patients With Acute Coronary Syndromes: The PROSPECT Study

Sex Differences in the Visual-Functional Mismatch Between Coronary Angiography or Intravascular Ultrasound Versus Fractional Flow Reserve

Background: Accurate segmentation of the coronary arteries with intravascular ultrasound (IVUS) is important to optimize coronary stent implantation. Recently, deep learning (DL) methods have been proposed to develop automatic IVUS segmentation. However, most of those have been limited to the lumen and vessel segmentation, not applied to segmenting stent dimension. Hence, this study aimed to develop a DL method for automatic IVUS segmentation of stent area in addition to lumen and vessel area. Methods: This study included a total of 45449 images from 1576 (40- or 45-MHz) IVUS pullback runs. The datasets were randomly split into training, validation, and test datasets (0.7:0.15:0.15). After constructing the convolutional neural network to segment IVUS images using the training and validation datasets, we evaluated the performance through the independent test dataset. Results: The DL-based segmentation correlated well with the expert-analysed segmentation with a mean intersection over union of 0.84, the correlation coefficient of 0.99, 0.98, and 0.99 and the mean difference of -0.04±0.05, -0.20±0.11 and -0.12±0.06 mm^2 for lumen, vessel and stent area derived from respectively. Conclusion: This automated DL-based IVUS segmentation of stent area in conjunction with lumen and vessel area showed good agreement with manual segmentation by experts, supporting the feasibility of artificial-intelligence-assisted IVUS assessment and strategy-making in patients undergoing coronary stent implantation.