Artificial Intelligence in Cardiovascular CT and MR Imaging

Training and assessing convolutional neural network performance in automatic vascular segmentation using Ga-68 DOTATATE PET/CT

To evaluate a convolutional neural network's performance (nnU-Net) in the assessment of vascular contours, calcification and PET tracer activity using Ga-68 DOTATATE PET/CT. Patients who underwent Ga-68 DOTATATE PET/CT imaging over a 12-month period for neuroendocrine investigation were included. Manual cardiac and aortic segmentations were performed by an experienced observer. Scans were randomly allocated in ratio 64:16:20 for training, validation and testing of the nnU-Net model. PET tracer uptake and calcium scoring were compared between segmentation methods and different observers. 116 patients (53.5% female) with a median age of 64.5 years (range 23-79) were included. There were strong, positive correlations between all segmentations (mostly r > 0.98). There were no significant differences between manual and AI segmentation of SUVmean for global cardiac (mean ± SD 0.71 ± 0.22 vs. 0.71 ± 0.22; mean diff 0.001 ± 0.008, p > 0.05), ascending aorta (mean ± SD 0.44 ± 0.14 vs. 0.44 ± 0.14; mean diff 0.002 ± 0.01, p > 0.05), aortic arch (mean ± SD 0.44 ± 0.10 vs. 0.43 ± 0.10; mean diff 0.008 ± 0.16, p > 0.05) and descending aorta (mean ± SD < 0.001; 0.58 ± 0.12 vs. 0.57 ± 0.12; mean diff 0.01 ± 0.03, p > 0.05) contours. There was excellent agreement between the majority of manual and AI segmentation measures (r ≥ 0.80) and in all vascular contour calcium scores. Compared with the manual segmentation approach, the CNN required a significantly lower workflow time. AI segmentation of vascular contours using nnU-Net resulted in very similar measures of PET tracer uptake and vascular calcification when compared to an experienced observer and significantly reduced workflow time.

Imaging biomarkers in cardiac CT: moving beyond simple coronary anatomical assessment

Cardiac computed tomography angiography (CCTA) is considered the standard non-invasive tool to rule-out obstructive coronary artery disease (CAD). Moreover, several imaging biomarkers have been developed on cardiac-CT imaging to assess global CAD severity and atherosclerotic burden, including coronary calcium scoring, the segment involvement score, segment stenosis score and the Leaman-score. Myocardial perfusion imaging enables the diagnosis of myocardial ischemia and microvascular damage, and the CT-based fractional flow reserve quantification allows to evaluate non-invasively hemodynamic impact of the coronary stenosis. The texture and density of the epicardial and perivascular adipose tissue, the hypodense plaque burden, the radiomic phenotyping of coronary plaques or the fat radiomic profile are novel CT imaging features emerging as biomarkers of inflammation and plaque instability, which may implement the risk stratification strategies. The ability to perform myocardial tissue characterization by extracellular volume fraction and radiomic features appears promising in predicting arrhythmogenic risk and cardiovascular events. New imaging biomarkers are expanding the potential of cardiac CT for phenotyping the individual profile of CAD involvement and opening new frontiers for the practice of more personalized medicine.

Contrast-enhanced cardiac MRI is superior to non-contrast mapping to predict left ventricular remodeling at 6 months after acute myocardial infarction

OBJECTIVES

Parametric mapping constitutes a novel cardiac magnetic resonance (CMR) technique enabling quantitative assessment of pathologic alterations of left ventricular (LV) myocardium. This study aimed to investigate the clinical utility of mapping techniques with and without contrast agent compared to standard CMR to predict adverse LV remodeling following acute myocardial infarction (AMI).

MATERIALS AND METHODS

A post hoc analysis was performed on sixty-four consecutively enrolled patients (57 ± 12 years, 54 men) with first-time reperfused AMI. Baseline CMR was obtained at 8 ± 5 days post-AMI, and follow-up CMR at 6 ± 1.4 months. T1/T2 mapping, T2-weighted, and late gadolinium enhancement (LGE) acquisitions were performed at baseline and cine imaging was used to determine adverse LV remodeling, defined as end-diastolic volume increase by 20% at 6 months.

RESULTS

A total of 11 (17%) patients developed adverse LV remodeling. At baseline, patients with LV remodeling showed larger edema (30 ± 11 vs. 22 ± 10%LV; p < 0.05), infarct size (24 ± 11 vs. 14 ± 8%LV; p < 0.001), extracellular volume (ECVinfarct; 63 ± 12 vs. 47 ± 11%; p < 0.001), and native T2infarct (95 ± 16 vs. 78 ± 17 ms; p < 0.01). ECVinfarct and infarct size by LGE were the best predictors of LV remodeling with areas under the curve (AUCs) of 0.843 and 0.789, respectively (all p < 0.01). Native T1infarct had the lowest AUC of 0.549 (p = 0.668) and was inferior to edema size by T2-weighted imaging (AUC = 0.720; p < 0.05) and native T2infarct (AUC = 0.766; p < 0.01).

CONCLUSION

In this study, ECVinfarct and infarct size by LGE were the best predictors for the development of LV remodeling within 6 months after AMI, with a better discriminative performance than non-contrast mapping CMR.

CLINICAL RELEVANCE STATEMENT

This study demonstrates the predictive value of contrast-enhanced and non-contrast as well as conventional and novel CMR techniques for the development of LV remodeling following AMI, which might help define precise CMR endpoints in experimental and clinical myocardial infarction trials.

KEY POINTS

• Multiparametric CMR provides insights into left ventricular remodeling at 6 months following an acute myocardial infarction. • Extracellular volume fraction and infarct size are the best predictors for adverse left ventricular remodeling. • Contrast-enhanced T1 mapping has a better predictive performance than non-contrast standard CMR and T1/T2 mapping.

Real-world approach to comprehensive artificial intelligence-aided CT evaluation of coronary artery disease in 530 patients: A retrospective study

Purpose

Recent guidelines provide broader support for the use of less invasive imaging modalities for the evaluation of patients with stable chest pain. Coronary CT angiography (CCTA) uses increasingly sophisticated techniques to improve evaluation of coronary lesions. The purpose of this study is to describe one center's experience implementing AI-assisted advanced imaging techniques to diagnose coronary artery disease.

Materials & methods

Retrospective study of patients who had AI-assisted CCTA interpretation, including a subgroup who underwent fractional flow reserve CT (FFR-CT) and invasive coronary angiography. Descriptive statistics summarized baseline characteristics and univariate statistics compared findings between groups of patients with and without anatomically and hemodynamically significant lesions based on FFR-CT. For patients who underwent invasive coronary angiography, concordance between CCTA and angiography was evaluated.

Results

Of 532 included patients, AI-assisted CCTA identified statistically significant difference in calcification scores, plaque types and total plaque volume between lesions <50% and ≥50% stenosis. CCTA results were mostly concordant with invasive coronary angiography. Importantly, we identified a subset of patients with less than 50% anatomical stenosis that demonstrated physiologically significant stenosis on FFR-CT and invasive coronary angiography.

Conclusions

AI-assisted CCTA and other advanced techniques are a tool to support high quality diagnostic assessment of coronary lesions in a clinical environment. Combined CCTA with FFRCT in mild to moderate coronary stenosis identifies patients with hemodynamically significant stenosis even when quantitative stenosis is <50%. Implementation of AI-assisted coronary CT angiography is feasible in a community hospital setting, but these technologies do not replace the need for expert review and clinical correlation.

MR imaging of primary benign cardiac tumors in the pediatric population

Primary cardiac tumors are rare in all ages, especially in children, with a reported prevalence range of 0.0017-0.28% in autopsy series. Due to their rarity, the diagnostic and therapeutic pathways reserved to them are usually described by single case reports, leading to the point where a common diagnostic protocol is imperative to obtain a differential diagnosis. The first diagnostic approach is done with transthoracic echocardiogram (TTE), due to its wide availability, low cost, absence of ionizing radiations and non-invasiveness. Several tumors are discovered incidentally and, in many cases, TTE is helpful to determine location, size and anatomical features, playing a key role in the differential diagnosis. In the last few years, cardiac magnetic resonance imaging (CMR) has had an increased role in the diagnostic pathway of pediatric cardiac masses, due to its high accuracy in characterizing mass tissue properties (especially for soft tissue), and in detecting tumor size, extent, pericardial/pleural effusion, leading to the correct diagnosis, treatment and follow-up. Therefore, nowadays, several consensus statements consider CMR as a leading imaging technique, thanks to its non-invasive tissue characterization, without the use of ionizing radiation, in an unrestricted field of view. As suggested by the most recent literature, the pediatric protocol is not so different from the adult one, adapted to the size and cardiac frequency of the patient, sometimes requiring special conditions such as free-breathing sequences and/or sedation or general anesthesia in non-cooperating patients.

Anomalous venoatrial connections - CT and MRI assessment

Abnormal venous atrial (VA) connections present a congenital heart disease (CHD) challenge for pediatric cardiologists. Fully anatomical evaluation is very difficult in prenatal and perinatal follow-up, but it has a profound impact on surgical correction and outcome. The echocardiogram is first-line imaging and represents the gold standard tool for simple abnormal VA connection. CT and MRI are mandatory for more complex heart disease and "nightmare cases". 3D post-processing of volumetric CT and MRI acquisition helps to clarify anatomical relationships and allows for the creation of 3D printing models that can become crucial in customizing surgical strategy. Our article describes a ten-year (2013-2022) tertiary referral CHD center of abnormal AV connections investigated with CT and MRI, illustrating most of these complex diseases with the help of volume rendering (VR) or multiplanar reconstructions (MPR). The nightmarish cases will also be addressed due to the complex cardiovascular arrangement that requires a challenging surgical solution for correction along with the post-surgical complications.

Automatic Classification of Magnetic Resonance Histology of Peripheral Arterial Chronic Total Occlusions Using a Variational Autoencoder: A Feasibility Study

The novel approach of our study consists in adapting and in evaluating a custom-made variational autoencoder (VAE) using two-dimensional (2D) convolutional neural networks (CNNs) on magnetic resonance imaging (MRI) images for differentiate soft vs. hard plaque components in peripheral arterial disease (PAD). Five amputated lower extremities were imaged at a clinical ultra-high field 7 Tesla MRI. Ultrashort echo time (UTE), T1-weighted (T1w) and T2-weighted (T2w) datasets were acquired. Multiplanar reconstruction (MPR) images were obtained from one lesion per limb. Images were aligned to each other and pseudo-color red-green-blue images were created. Four areas in latent space were defined corresponding to the sorted images reconstructed by the VAE. Images were classified from their position in latent space and scored using tissue score (TS) as following: (1) lumen patent, TS:0; (2) partially patent, TS:1; (3) mostly occluded with soft tissue, TS:3; (4) mostly occluded with hard tissue, TS:5. Average and relative percentage of TS was calculated per lesion defined as the sum of the tissue score for each image divided by the total number of images. In total, 2390 MPR reconstructed images were included in the analysis. Relative percentage of average tissue score varied from only patent (lesion #1) to presence of all four classes. Lesions #2, #3 and #5 were classified to contain tissues except mostly occluded with hard tissue while lesion #4 contained all (ranges (I): 0.2-100%, (II): 46.3-75.9%, (III): 18-33.5%, (IV): 20%). Training the VAE was successful as images with soft/hard tissues in PAD lesions were satisfactory separated in latent space. Using VAE may assist in rapid classification of MRI histology images acquired in a clinical setup for facilitating endovascular procedures.