CT imaging features of carotid artery plaque vulnerability

Clinical Application of Dark-blood Imaging in Head and Neck CT Angiography: Effect on Image Quality and Plaque Visibility

RATIONALE AND OBJECTIVES

The aim of this study was to explore the potential of a newly developed dark-blood imaging technique to improve image quality and plaque visibility in head and neck computed tomography (CT) angiography.

MATERIALS AND METHODS

Patients who underwent triphasic head and neck CT angiography scans from August 2021 to March 2023 were retrospectively enrolled (mean age 67.23 ± 10.81 [SD] years, range 43-85 years, 64.7% male). The CT protocol consists of pre-contrast, arterial and delayed phases. Dark-blood images were postprocessed with the contrast-enhancement boost (CE-boost) technique. The quantitative assessment involved evaluating the CT value, image noise, contrast-to-noise ratio (CNR), and signal-to-noise ratio (SNR) of calcified plaque and non-calcified plaque. The plaque CNR relative to the vessel lumen (CNRplaque-lumen), vessel wall (CNRplaque-wall), and adjacent muscle (CNRplaque-muscle) was respectively calculated. Two experienced radiologists independently evaluated the CT images (5, best; 1, worst) by four characteristics including calcified plaque visibility, non-calcified plaque visibility, diagnostic confidence, and overall image quality. Inter-rater variability was also evaluated. The artery stenosis rate and plaque burden on dark-blood images were measured and compared with arterial phases. The intraclass correlation coefficient (ICC) was used for consistency analysis. The diagnostic accuracy of dark-blood images for the stenosis rate was evaluated by the area under the curve (AUC).

RESULTS

A total of 43 patients with 54 calcified plaques and 34 non-calcified plaques were assessed in this study. When compared with pre-contrast and delayed phase, dark-blood images yielded significantly higher CNRplaque-lumen and CNRplaque-muscle of calcified (219.79 ± 159.20 and 181.23 ± 112.12, respectively) and non-calcified (30.30 ± 29.11 and 6.28 ± 4.75, respectively) plaques (all p < 0.001). Calcified plaque SNR of dark-blood showed equal or slightly lower than other phases (p > 0.05 or p = 0.02). A major increase was observed in the non-calcified plaque SNR of dark-blood compared to the arterial phase (5.56 ± 3.71 vs. 4.23 ± 3.56, p = 0.02), although there were no apparent differences compared to pre-contrast and delayed phases (p > 0.05). In subjective analyzes, the calcified plaque visibility (4.99 ± 0.07), non-calcified plaque visibility (4.62 ± 0.48), overall image quality (4.81 ± 0.34), and diagnostic confidence (4.74 ± 0.36) in dark-blood images dominated the highest scores (p < 0.001). The subjective scores of radiologists exhibited good consistency (all kappa value>0.7). The dark-blood image and the arterial phase image exhibited good consistency in identifying the stenosis rate (p < 0.001). In the evaluation of plaque burden, the interobserver agreement for dark-blood images was higher compared to arterial phase images (ICC = 0.870 vs. 0.729).

CONCLUSIONS

Compared to conventional triphasic head and neck CT angiography, the CE-boost derived dark-blood imaging demonstrated a significant improvement in image quality and visibility for both calcified and non-calcified plaque assessment.

A novel rabbit model of atherosclerotic vulnerable plaque established by cryofluid-induced endothelial injury

Acute thrombosis secondary to atherosclerotic plaque rupture is the main cause of acute cardiac and cerebral ischemia. An animal model of unstable atherosclerotic plaques is highly important for investigating the mechanism of plaque rupture and thrombosis. However, current animal models involve complex operations, are costly, and have plaque morphologies that are different from those of humans. We aimed to establish a simple animal model of vulnerable plaques similar to those of humans. Rabbits were randomly divided into three groups. Group A was given a normal formula diet for 13 weeks. Group C underwent surgery on the intima of the right carotid artery with - 80 °C cryofluid-induced injury after 1 week of a high-fat diet and further feeding a 12-week high-fat diet. Group B underwent the same procedure as Group C but without the - 80 °C cryofluid. Serum lipid levels were detected via ELISA. The plaque morphology, stability and degree of stenosis were evaluated through hematoxylin-eosin (HE) staining, Masson trichrome staining, Elastica van Gieson staining (EVG), and oil red O staining. Macrophages and inflammatory factors in the plaques were assessed via immunohistochemical analysis. The serum low-density lipoprotein cholesterol (LDL-C), triglyceride (TG), and total cholesterol (TC) levels in groups B and C were significantly greater than those in group A. No plaque formation was observed in group A. The plaques in group B were very small. In group C, obvious plaques were observed in the blood vessels, and the plaques exhibited a thin fibrous cap, a large lipid core, and partially visible neovascularization, which is consistent with the characteristics of vulnerable plaques. In the plaques of group C, a large number of macrophages were present, and matrix metalloproteinase 9 (MMP-9) and lectin-like oxidized LDL receptor 1 (LOX-1) were abundantly expressed. We successfully established a rabbit model of vulnerable carotid plaque similar to that of humans through the combination of cryofluid-induced endothelial injury and a high-fat diet, which is feasible and cost effective.

Update on cardiac imaging: A critical analysis

Imaging is instrumental in diagnosing and directing the management of atherosclerosis. In 1958 the first diagnostic coronary angiography (CA) was performed, and since then further development has led to new methods such as coronary CT angiography (CTA), optical coherence tomography (OCT), positron tomography (PET), and intravascular ultrasound (IVUS). Currently, CA remains powerful for visualizing coronary arteries; however, recent studies show the benefits of using other non-invasive techniques. This review identifies optimum imaging techniques for diagnosing and monitoring plaque stability. This becomes even direr now, given the rapidly rising incidence of atherosclerosis in society today. Many acute coronary events, including acute myocardial infarctions and sudden deaths, are attributable to plaque rupture. Although fatal, these events can be preventable. We discuss the factors affecting plaque integrity, such as increased inflammation, medications like statins, and increased lipid content. Some of these precipitating factors are identifiable through imaging. However, we also highlight significant complications arising in some modalities; in CA this can include ventricular arrhythmia and even death. Extending this, we elucidated from the literature that risk can also vary based on the location of arteries and their plaques. Promisingly, there are less invasive methods being trialled for assessing plaque stability, such as Cardiac Magnetic Resonance Imaging (CMR), which is already in use for other cardiac diseases like cardiomyopathies. Therefore, future research focusing on using imaging modalities in conjunction may be sensible, to bridge between the effectiveness of modalities, at the expense of increased complications, and vice versa.

Clinical risk factors and features on computed tomography angiography in high-risk carotid artery plaque in patients with type 2 diabetes

BACKGROUND

High-risk carotid artery plaque (HPR) is associated with a markedly increased risk of ischemic stroke. The aims of this study were: 1) to examine the prevalence of HRP in a cohort of asymptomatic adults with type 2 diabetes (T2D); 2) to investigate the relationship between HRP, established cardiovascular risk factors and computed tomography angiography (CTA) profile; and 3) to assess whether the presence of HRP is associated with an increased risk of major adverse cardiovascular events (MACE).

METHODS

This was a retrospective cohort study of T2D asymptomatic patients who underwent carotid endarterectomy (CEA) from January 2018 to July 2021. The carotid atherosclerotic plaque (CAP) was assessed for the presence of ulceration, the presence of lipids, fibrosis, thrombotic deposits, hemorrhage, neovascularization, and inflammation. A CAP presenting at least five of these histological features was defined as a HRP (Group A); in all other cases it was defined as a mild to moderate heterogeneous plaque and no-HRP (Group B). CTA features included the presence of rim sign consisting of thin peripheral adventitial calcification (<2 mm) and internal soft plaque (≥2 mm), NASCET percent diameter stenosis, maximum plaque thickness, ulceration, calcification, and intraluminal thrombus were recorded. Binary logistic regression with Uni- and Multivariate was used to evaluate possible predictors for HRP while multivariable Cox Proportional Hazards was used to assess independent predictors for MACE.

RESULTS

One hundred eighty-five asymptomatic patients (mean age 73±8 years, 131 men), undergoing carotid endarterectomy, were included. Of these, 124 (67%) had HRP, and the 61 (33%) did not. Diabetic complications (OR 2.4, 95% CI: 1.1-5.1, P=0.01), NASCET stenosis ≥75% (OR 2.4, 95% CI: 1.2-3.7, P=0.02) and carotid RIM sign (OR 4.3, 95% CI: 3.9-7.3, P<0.001) were independently associated with HRP. However, HRP was not associated with a higher risk of MACE (freedom from MACE at 5 years: HRP 83.4% vs. non HRP 87.8%, P=0.72) or a reduction of survival (5-year survival estimates: HRP 96.4% vs. non HRP: 94.6%, P=0.76).

CONCLUSIONS

A high prevalence of HRP (67%) was observed in asymptomatic and elderly T2D patients. Independent predictors of HRP were diabetic complications, NASCET stenosis ≥75% and carotid RIM sign (OR 4.3, 95% CI: 3.9-7.3, P<0.001). HRP was not associated with an increased risk of MACE during a mean follow-up of 39±24 years.

Risk of new cerebral ischemic events in patients with symptomatic internal carotid artery stenosis while awaiting carotid stent placement

BACKGROUND AND PURPOSE

Although there is an emphasis on performing carotid artery stent (CAS) placement within 2 weeks after index event of transient ischemic attack (TIA) or minor stroke in patients with significant extracranial internal carotid artery (ICA) stenosis, the risks and characteristics of recurrent cerebral ischemic event while waiting for CAS placement are not well defined.

METHOD

We analyzed patients admitted to our institution over a 45-month period with symptomatic extracranial ICA stenosis. We identified any new cerebral ischemic events that occurred between index cerebral or retinal ischemic event and CAS placement and categorized them as TIA and minor or major ischemic strokes. We calculated the risk of new ipsilateral cerebral ischemic events between index cerebral or retinal ischemic event and CAS placement.

RESULTS

The mean age of 131 patients analyzed was 67 years (range: 47-94 years; 92 were men), and 94 and 37 patients had 70%-99% and 50%-69% severity of stenosis, respectively. The mean and median time intervals between index cerebral or retinal ischemic event and CAS performance were 28 (standard deviation [SD] 30) and 7 (interquartile range 33) days, respectively. A total of 9 of 131 patients (6.9%, 95% confidence interval 2.5%-11.2%) experienced new cerebral ischemic events over 3637 patient days of observation. The risk of new ipsilateral cerebral ischemic events was 2.5 per 1000 patient days of observation.

CONCLUSION

We estimated the risk of new ipsilateral cerebral ischemic events in patients with ICA stenosis ≥50% in severity while waiting for CAS placement to guide appropriate timing of procedure.

Contrast-Enhanced Ultrasound Feasibility in Assessing Carotid Plaque Vulnerability-Narrative Review

The risk assessment for carotid atherosclerotic lesions involves not only determining the degree of stenosis but also plaque morphology and its composition. Recently, carotid contrast-enhanced ultrasound (CEUS) has gained importance for evaluating vulnerable plaques. This review explores CEUS's utility in detecting carotid plaque surface irregularities and ulcerations as well as intraplaque neovascularization and its alignment with histology. Initial indications suggest that CEUS might have the potential to anticipate cerebrovascular incidents. Nevertheless, there is a need for extensive, multicenter prospective studies that explore the relationships between CEUS observations and patient clinical outcomes in cases of carotid atherosclerotic disease.

Ensemble Deep Learning Derived from Transfer Learning for Classification of COVID-19 Patients on Hybrid Deep-Learning-Based Lung Segmentation: A Data Augmentation and Balancing Framework

BACKGROUND AND MOTIVATION

Lung computed tomography (CT) techniques are high-resolution and are well adopted in the intensive care unit (ICU) for COVID-19 disease control classification. Most artificial intelligence (AI) systems do not undergo generalization and are typically overfitted. Such trained AI systems are not practical for clinical settings and therefore do not give accurate results when executed on unseen data sets. We hypothesize that ensemble deep learning (EDL) is superior to deep transfer learning (TL) in both non-augmented and augmented frameworks.

METHODOLOGY

The system consists of a cascade of quality control, ResNet-UNet-based hybrid deep learning for lung segmentation, and seven models using TL-based classification followed by five types of EDL's. To prove our hypothesis, five different kinds of data combinations (DC) were designed using a combination of two multicenter cohorts-Croatia (80 COVID) and Italy (72 COVID and 30 controls)-leading to 12,000 CT slices. As part of generalization, the system was tested on unseen data and statistically tested for reliability/stability.

RESULTS

Using the K5 (80:20) cross-validation protocol on the balanced and augmented dataset, the five DC datasets improved TL mean accuracy by 3.32%, 6.56%, 12.96%, 47.1%, and 2.78%, respectively. The five EDL systems showed improvements in accuracy of 2.12%, 5.78%, 6.72%, 32.05%, and 2.40%, thus validating our hypothesis. All statistical tests proved positive for reliability and stability.

CONCLUSION

EDL showed superior performance to TL systems for both (a) unbalanced and unaugmented and (b) balanced and augmented datasets for both (i) seen and (ii) unseen paradigms, validating both our hypotheses.

Pearls and Pitfalls of Carotid Artery Imaging: Ultrasound, Computed Tomography Angiography, and MR Imaging

Stroke represents a major cause of morbidity and mortality worldwide with carotid atherosclerosis responsible for a large proportion of ischemic strokes. Given the high burden of the disease , early diagnosis and optimal secondary prevention are essential elements in clinical practice. For a long time, the degree of stenosis had been considered the parameter to judge the severity of carotid atherosclerosis. Over the last 30 years, literature has shifted attention from stenosis to structural characteristics of atherosclerotic lesion, eventually leading to the "vulnerable plaque" model. These "vulnerable plaques" frequently demonstrate high-risk imaging features that can be assessed by various non-invasive imaging modalities.

Choice of imaging to evaluate carotid stenosis and guide management

Carotid artery disease is a cause of ischemic stroke and is associated with cognitive decline. Besides the evaluation of the degree of stenosis, it is also crucial to assess the morphology of the atherosclerotic plaque, for a prompt and accurate diagnosis, and to make the best decision for the patient. On top of noninvasive duplex ultrasound (DUS) and invasive digital subtraction angiography (DSA), compute tomography angiography (CTA) and magnetic resonance angiography (MRA) are often used effectively as noninvasive imaging tools to study carotid stenoses. This review describes the fundamental characteristics of carotid artery plaques, and how they can be best evaluated with currently available imaging methods.

The utility of ultrasound and computed tomography in the assessment of carotid artery plaque vulnerability-A mini review

As the burden of cardiovascular and cerebrovascular events continues to increase, emerging evidence supports the concept of plaque vulnerability as a strong marker of plaque rupture, and embolization. Qualitative assessment of the plaque can identify the degree of plaque instability. Ultrasound and computed tomography (CT) have emerged as safe and accurate techniques for the assessment of plaque vulnerability. Plaque features including but not limited to surface ulceration, large lipid core, thin fibrous cap (FC), intraplaque neovascularization and hemorrhage can be assessed and are linked to plaque instability.

Characteristics of Calcification and Their Association with Carotid Plaque Vulnerability

BACKGROUND

Carotid plaque vulnerability is one of the important features for evaluating the risk of subsequent ischemic stroke. Although magnetic resonance imaging (MRI) is the gold standard modality for evaluating plaque vulnerability, some patients cannot undergo MRI because of physical or economic issues. Computed tomography (CT) is more readily available. The purpose of this study was to establish a new category of calcification on CT and to assess its usefulness for detecting vulnerable plaque.

MATERIALS AND METHODS

We retrospectively evaluated consecutive patients who underwent plaque imaging using CT and MRI before carotid revascularization at our institute. Calcifications were classified into 4 types according to the new calcium classification. The patients were divided into 2 groups, the double layer sign (DLS)-positive group and the DLS-negative group. Signal intensity ratio (SIR) of carotid plaque was measured on MRI for evaluating plaque vulnerability and compared between type of calcification and SIR.

RESULTS

Among the 132 patients evaluated, 50 patients (62.5%) in DLS positive group and 16 patients (30.8%) in DLS negative group had calcification with vulnerable plaque (SIR > 1.47) (P < 0.01). Substantial interobserver agreement of type of calcification was observed (kappa, 0.79; P < 0.01). Multivariate analysis showed that DLS (odds ratio 3.03; 95% confidence interval 1.35-6.8; P < 0.01) and male sex (odds ratio 3.15; 95% confidence interval 1.02-9.68; P = 0.04) were independent predictors of vulnerable plaque.

CONCLUSIONS

DLS in our new classification of calcification on CT reliably detects vulnerable plaque and could thus be used in patients who cannot undergo MRI.

Comparative assessment of the signs of instability of atherosclerotic plaques in the carotid arteries in elderly patients with acute coronary syndrome with duplex scanning and computed tomography angiography

Aim. To study the signs of instability of atherosclerotic plaques (ASP) in the carotid arteries in elderly patients with acute coronary syndrome (ACS) in the comparative aspect of duplex ultrasound scanning (DUS) and computed tomography angiography (CT).

Material and methods. 27 patients with ACS (75 years and over) were included in the study. The signs of instability of ASP were assessed according to DUS and CT.

Results. The signs of instability of ASP according to DUS were detected in 85,7%, according to CT — in 84,6%. The following signs were detected with DUS and CT: the presence of irregular plaque surface including signs of ulceration — 6,4 and 11,6% (p=0,021), positive remodeling — 3,8 and 3,8% (p=0,998), signs of local calcification — 23 and 25,9% (p=0,536), heterogenous structure — 55,1 and 46,8% (p=0,045), hypoechogenic component and low-density areas — 11,5 and 11,6% (p= 0,998). The correlation analysis showed high comparability of DUS and CT: irregular plaque surface with ulceration (K=0,624, p=0,02), positive remodeling (K=1, p<0,001), calcification (K=0,858, p<0,001), heterogenous structure (K=0,754, p<0,001), the presence of hypoechogenic component and low-density areas (K=1, p<0,001).

Conclusion. The study of elderly patients with ACS found high comparability of DUS and CT in the definition of the signs of instability of ASP in the carotid arteries. It is possible to use DUS as a routine method for assessing carotid atherosclerosis in patients of this group, which can reduce the risk of complications during CT, shorten the examination time, and minimize economic costs.

Identification Markers of Carotid Vulnerable Plaques: An Update

Vulnerable plaques have been a hot topic in the field of stroke and carotid atherosclerosis. Currently, risk stratification and intervention of carotid plaques are guided by the degree of luminal stenosis. Recently, it has been recognized that the vulnerability of plaques may contribute to the risk of stroke. Some classical interventions, such as carotid endarterectomy, significantly reduce the risk of stroke in symptomatic patients with severe carotid stenosis, while for asymptomatic patients, clinically silent plaques with rupture tendency may expose them to the risk of cerebrovascular events. Early identification of vulnerable plaques contributes to lowering the risk of cerebrovascular events. Previously, the identification of vulnerable plaques was commonly based on imaging technologies at the macroscopic level. Recently, some microscopic molecules pertaining to vulnerable plaques have emerged, and could be potential biomarkers or therapeutic targets. This review aimed to update the previous summarization of vulnerable plaques and identify vulnerable plaques at the microscopic and macroscopic levels.

Multicenter Study on COVID-19 Lung Computed Tomography Segmentation with varying Glass Ground Opacities using Unseen Deep Learning Artificial Intelligence Paradigms: COVLIAS 1.0 Validation

Variations in COVID-19 lesions such as glass ground opacities (GGO), consolidations, and crazy paving can compromise the ability of solo-deep learning (SDL) or hybrid-deep learning (HDL) artificial intelligence (AI) models in predicting automated COVID-19 lung segmentation in Computed Tomography (CT) from unseen data leading to poor clinical manifestations. As the first study of its kind, “COVLIAS 1.0-Unseen” proves two hypotheses, (i) contrast adjustment is vital for AI, and (ii) HDL is superior to SDL. In a multicenter study, 10,000 CT slices were collected from 72 Italian (ITA) patients with low-GGO, and 80 Croatian (CRO) patients with high-GGO. Hounsfield Units (HU) were automatically adjusted to train the AI models and predict from test data, leading to four combinations—two Unseen sets: (i) train-CRO:test-ITA, (ii) train-ITA:test-CRO, and two Seen sets: (iii) train-CRO:test-CRO, (iv) train-ITA:test-ITA. COVILAS used three SDL models: PSPNet, SegNet, UNet and six HDL models: VGG-PSPNet, VGG-SegNet, VGG-UNet, ResNet-PSPNet, ResNet-SegNet, and ResNet-UNet. Two trained, blinded senior radiologists conducted ground truth annotations. Five types of performance metrics were used to validate COVLIAS 1.0-Unseen which was further benchmarked against MedSeg, an open-source web-based system. After HU adjustment for DS and JI, HDL (Unseen AI) > SDL (Unseen AI) by 4% and 5%, respectively. For CC, HDL (Unseen AI) > SDL (Unseen AI) by 6%. The COVLIAS-MedSeg difference was < 5%, meeting regulatory guidelines.Unseen AI was successfully demonstrated using automated HU adjustment. HDL was found to be superior to SDL.

A biomathematical model of atherosclerosis in mice

Atherosclerosis is one of the leading causes of death worldwide. Biomathematical modelling of the underlying disease and therapy processes might be a useful aid to develop and improve preventive and treatment concepts of atherosclerosis. We here propose a biomathematical model of murine atherosclerosis under different diet and treatment conditions including lipid modulating compound and antibiotics. The model is derived by translating known biological mechanisms into ordinary differential equations and by assuming appropriate response kinetics to the applied interventions. We explicitly describe the dynamics of relevant immune cells and lipid species in atherosclerotic lesions including the degree of blood vessel occlusion due to growing plaques. Unknown model parameters were determined by fitting the predictions of model simulations to time series data derived from mice experiments. Parameter fittings resulted in a good agreement of model and data for all 13 experimental scenarios considered. The model can be used to predict the outcome of alternative treatment schedules of combined antibiotic, immune modulating, and lipid lowering agents under high fat or normal diet. We conclude that we established a comprehensive biomathematical model of atherosclerosis in mice. We aim to validate the model on the basis of further experimental data.

Eight pruning deep learning models for low storage and high-speed COVID-19 computed tomography lung segmentation and heatmap-based lesion localization: A multicenter study using COVLIAS 2.0

BACKGROUND

COVLIAS 1.0: an automated lung segmentation was designed for COVID-19 diagnosis. It has issues related to storage space and speed. This study shows that COVLIAS 2.0 uses pruned AI (PAI) networks for improving both storage and speed, wiliest high performance on lung segmentation and lesion localization.

METHOD

ology: The proposed study uses multicenter ∼9,000 CT slices from two different nations, namely, CroMed from Croatia (80 patients, experimental data), and NovMed from Italy (72 patients, validation data). We hypothesize that by using pruning and evolutionary optimization algorithms, the size of the AI models can be reduced significantly, ensuring optimal performance. Eight different pruning techniques (i) differential evolution (DE), (ii) genetic algorithm (GA), (iii) particle swarm optimization algorithm (PSO), and (iv) whale optimization algorithm (WO) in two deep learning frameworks (i) Fully connected network (FCN) and (ii) SegNet were designed. COVLIAS 2.0 was validated using "Unseen NovMed" and benchmarked against MedSeg. Statistical tests for stability and reliability were also conducted.

RESULTS

Pruning algorithms (i) FCN-DE, (ii) FCN-GA, (iii) FCN-PSO, and (iv) FCN-WO showed improvement in storage by 92.4%, 95.3%, 98.7%, and 99.8% respectively when compared against solo FCN, and (v) SegNet-DE, (vi) SegNet-GA, (vii) SegNet-PSO, and (viii) SegNet-WO showed improvement by 97.1%, 97.9%, 98.8%, and 99.2% respectively when compared against solo SegNet. AUC > 0.94 (p < 0.0001) on CroMed and > 0.86 (p < 0.0001) on NovMed data set for all eight EA model. PAI <0.25 s per image. DenseNet-121-based Grad-CAM heatmaps showed validation on glass ground opacity lesions.

CONCLUSIONS

Eight PAI networks that were successfully validated are five times faster, storage efficient, and could be used in clinical settings.

COVLIAS 2.0-cXAI: Cloud-Based Explainable Deep Learning System for COVID-19 Lesion Localization in Computed Tomography Scans

Background: The previous COVID-19 lung diagnosis system lacks both scientific validation and the role of explainable artificial intelligence (AI) for understanding lesion localization. This study presents a cloud-based explainable AI, the “COVLIAS 2.0-cXAI” system using four kinds of class activation maps (CAM) models. Methodology: Our cohort consisted of ~6000 CT slices from two sources (Croatia, 80 COVID-19 patients and Italy, 15 control patients). COVLIAS 2.0-cXAI design consisted of three stages: (i) automated lung segmentation using hybrid deep learning ResNet-UNet model by automatic adjustment of Hounsfield units, hyperparameter optimization, and parallel and distributed training, (ii) classification using three kinds of DenseNet (DN) models (DN-121, DN-169, DN-201), and (iii) validation using four kinds of CAM visualization techniques: gradient-weighted class activation mapping (Grad-CAM), Grad-CAM++, score-weighted CAM (Score-CAM), and FasterScore-CAM. The COVLIAS 2.0-cXAI was validated by three trained senior radiologists for its stability and reliability. The Friedman test was also performed on the scores of the three radiologists. Results: The ResNet-UNet segmentation model resulted in dice similarity of 0.96, Jaccard index of 0.93, a correlation coefficient of 0.99, with a figure-of-merit of 95.99%, while the classifier accuracies for the three DN nets (DN-121, DN-169, and DN-201) were 98%, 98%, and 99% with a loss of ~0.003, ~0.0025, and ~0.002 using 50 epochs, respectively. The mean AUC for all three DN models was 0.99 (p < 0.0001). The COVLIAS 2.0-cXAI showed 80% scans for mean alignment index (MAI) between heatmaps and gold standard, a score of four out of five, establishing the system for clinical settings. Conclusions: The COVLIAS 2.0-cXAI successfully showed a cloud-based explainable AI system for lesion localization in lung CT scans.

Reassessing the Carotid Artery Plaque "Rim Sign" on CTA: A New Analysis with Histopathologic Confirmation

BACKGROUND AND PURPOSE

The CTA "rim sign" has been proposed as an imaging marker of intraplaque hemorrhage in carotid plaques. This study sought to investigate such findings using histopathologic confirmation.

MATERIALS AND METHODS

Included patients had CTA neck imaging <1 year before carotid endarterectomy. On imaging, luminal stenosis and the presence of adventitial (<2-mm peripheral) and "bulky" (≥2-mm) calcifications, total plaque thickness, soft-tissue plaque thickness, calcification thickness, and the presence of ulcerations were assessed. The rim sign was defined as the presence of adventitial calcifications with internal soft-tissue plaque of ≥2 mm in maximum thickness. Carotid endarterectomy specimens were assessed for both the presence and the proportional makeup of lipid material, intraplaque hemorrhage, and calcification.

RESULTS

Sixty-seven patients were included. Twenty-three (34.3%) were women; the average age was 70.4 years. Thirty-eight (57.7%) plaques had a rim sign on imaging, with strong interobserver agreement (κ = 0.85). A lipid core was present in 64 (95.5%) plaques (average, 22.2% proportion of plaque composition); intraplaque hemorrhage was present in 52 (77.6%), making up, on average, 13.7% of the plaque composition. The rim sign was not associated with the presence of intraplaque hemorrhage (P = .11); however, it was associated with a greater proportion of intraplaque hemorrhage in a plaque (P = .049). The sensitivity and specificity of the rim sign for intraplaque hemorrhage were 61.5% and 60.0%, respectively.

CONCLUSIONS

The rim sign is not associated with the presence of intraplaque hemorrhage on histology. However, it is associated with a higher proportion of hemorrhage within a plaque and therefore may be a biomarker of more severe intraplaque hemorrhage, if present.

A hybrid deep learning paradigm for carotid plaque tissue characterization and its validation in multicenter cohorts using a supercomputer framework

BACKGROUND

Early and automated detection of carotid plaques prevents strokes, which are the second leading cause of death worldwide according to the World Health Organization. Artificial intelligence (AI) offers automated solutions for plaque tissue characterization. Recently, solo deep learning (SDL) models have been used, but they do not take advantage of the tandem connectivity offered by AI's hybrid nature. Therefore, this study explores the use of hybrid deep learning (HDL) models in a multicenter framework, making this study the first of its kind.

METHODS

We hypothesize that HDL techniques perform better than SDL and transfer learning (TL) techniques. We propose two kinds of HDL frameworks: (i) the fusion of two SDLs (Inception with ResNet) or (ii) 10 other kinds of tandem models that fuse SDL with ML. The system Atheromatic™ 2.0_HDL (AtheroPoint, CA, USA) was designed on an augmentation framework and three kinds of loss functions (cross-entropy, hinge, and mean-square-error) during training to determine the best optimization paradigm. These 11 combined HDL models were then benchmarked against one SDL model and five types of TL models; thus, this study considers a total of 17 AI models.

RESULTS

Among the 17 AI models, the best performing HDL system was that comprising CNN and decision tree (DT), as its accuracy and area-under-the-curve were 99.78 ± 1.05% and 0.99 (p<0.0001), respectively. These values are 6.4% and 3.2% better than those recorded for the SDL and TL models, respectively. We validated the performance of the HDL models with diagnostics odds ratio (DOR) and Cohen and Kappa statistics; here, HDL outperformed DL and TL by 23% and 7%, respectively. The online system ran in <2 s.

CONCLUSION

HDL is a fast, reliable, and effective tool for characterizing the carotid plaque for early stroke risk stratification.

Hybrid deep learning segmentation models for atherosclerotic plaque in internal carotid artery B-mode ultrasound

The automated and accurate carotid plaque segmentation in B-mode ultrasound (US) is an essential part of stroke risk stratification. Previous segmented methods used AtheroEdge™ 2.0 (AtheroPoint™, Roseville, CA) for the common carotid artery (CCA). This study focuses on automated plaque segmentation in the internal carotid artery (ICA) using solo deep learning (SDL) and hybrid deep learning (HDL) models. The methodology consists of a novel design of 10 types of SDL/HDL models (AtheroEdge™ 3.0 systems (AtheroPoint™, Roseville, CA) with a depth of four layers each. Five of the models use cross-entropy (CE)-loss, and the other five models use Dice similarity coefficient (DSC)-loss functions derived from UNet, UNet+, SegNet, SegNet-UNet, and SegNet-UNet+. The K10 protocol (Train:Test:90%:10%) was applied for all 10 models for training and predicting (segmenting) the plaque region, which was then quantified to compute the plaque area in mm². Further, the data augmentation effect was analyzed. The database consisted of 970 ICA B-mode US scans taken from 99 moderate to high-risk patients. Using the difference area threshold of 10 mm² between ground truth (GT) and artificial intelligence (AI), the area under the curve (AUC) values were 0.91, 0.911, 0.908, 0.905, and 0.898, all with a p-value of <0.001 (for CE-loss models) and 0.883, 0.889, 0.905, 0.889, and 0.907, all with a p-value of <0.001 (for DSC-loss models). The correlations between the AI-based plaque area and GT plaque area were 0.98, 0.96, 0.97, 0.98, and 0.97, all with a p-value of <0.001 (for CE-loss models) and 0.98, 0.98, 0.97, 0.98, and 0.98 (for DSC-loss models). Overall, the online system performs plaque segmentation in less than 1 s. We validate our hypothesis that HDL and SDL models demonstrate comparable performance. SegNet-UNet was the best-performing hybrid architecture.

Multimodality carotid plaque tissue characterization and classification in the artificial intelligence paradigm: a narrative review for stroke application

Cardiovascular disease (CVD) is one of the leading causes of morbidity and mortality in the United States of America and globally. Carotid arterial plaque, a cause and also a marker of such CVD, can be detected by various non-invasive imaging modalities such as magnetic resonance imaging (MRI), computer tomography (CT), and ultrasound (US). Characterization and classification of carotid plaque-type in these imaging modalities, especially into symptomatic and asymptomatic plaque, helps in the planning of carotid endarterectomy or stenting. It can be challenging to characterize plaque components due to (I) partial volume effect in magnetic resonance imaging (MRI) or (II) varying Hausdorff values in plaque regions in CT, and (III) attenuation of echoes reflected by the plaque during US causing acoustic shadowing. Artificial intelligence (AI) methods have become an indispensable part of healthcare and their applications to the non-invasive imaging technologies such as MRI, CT, and the US. In this narrative review, three main types of AI models (machine learning, deep learning, and transfer learning) are analyzed when applied to MRI, CT, and the US. A link between carotid plaque characteristics and the risk of coronary artery disease is presented. With regard to characterization, we review tools and techniques that use AI models to distinguish carotid plaque types based on signal processing and feature strengths. We conclude that AI-based solutions offer an accurate and robust path for tissue characterization and classification for carotid artery plaque imaging in all three imaging modalities. Due to cost, user-friendliness, and clinical effectiveness, AI in the US has dominated the most.