Radiomics Signatures of Cardiovascular Risk Factors in Cardiac MRI: Results From the UK Biobank

An interpretable radiomics-based machine learning model for predicting reverse left ventricular remodeling in STEMI patients using late gadolinium enhancement of myocardial scar

OBJECTIVES

To evaluate the added value of the late gadolinium enhancement (LGE)-scar radiomics features in predicting reverse left ventricular remodeling (r-LVR) in ST-segment elevation myocardial infarction (STEMI) patients using machine learning (ML).

MATERIALS AND METHODS

This retrospective study included 105 STEMI patients who underwent CMR within 7 days and 5 months post-percutaneous coronary intervention (PCI) on 1.5-T or 3.0-T MRI scanners (January 2014-2023). Radiomics features from LGE scar images and routine CMR markers were analyzed using a LightGBM model enhanced by Shapley Additive exPlanations (SHAP) for interpretability. Patients were divided into training (80) and test (25) sets. Three predictive models were developed: traditional CMR, LGE-scar radiomics, and a combined model integrating both. Model performance was assessed using ROC curves and AUC analysis.

RESULTS

In the training set, the traditional CMR model achieved an AUC of 0.745 (95% CI: 0.62-0.86), the LGE-scar radiomics model had an AUC of 0.712 (95% CI: 0.58-0.83), and the combined model showed the highest AUC of 0.754 (95% CI: 0.63-0.86). In the test set, the traditional CMR model's AUC decreased to 0.656 (95% CI: 0.42-0.88), while the LGE-scar radiomics model improved to 0.818 (95% CI: 0.59-1.00). The combined model achieved the highest AUC of 0.890 (95% CI: 0.75-1.00). SHAP analysis highlighted significant predictors such as infarct percentage of LV mass and wavelet-transformed texture features.

CONCLUSION

Integrating LGE scar radiomics features with traditional CMR parameters in a LightGBM model enhances predictive accuracy for r-LVR in STEMI patients, potentially improving patient stratification and treatment personalization.

KEY POINTS

Question Predicting r-LVR in STEMI patients remains challenging due to limitations in current imaging approaches. Findings Integrating LGE-scar radiomics and cardiac magnetic resonance markers in the LightGBM model significantly improves prediction accuracy for r-LVR. Clinical relevance This interpretable ML model enhances r-LVR prediction, supporting patient stratification and optimizing treatment strategies to improve patient outcomes.

Addressing Inter-reconstruction variability in multi-energy myocardial CT Radiomics: The Benefits of combat harmonization

RATIONALE AND OBJECTIVES

To investigate the effect of ComBat harmonization on the stability of myocardial radiomic features derived from multi-energy CT reconstructions.

MATERIALS AND METHODS

A retrospective study was conducted on 205 patients who underwent dual-energy chest CTA at a single center. The data was reconstructed into multiple spectral reconstructions (mixed energy simulating standard 120 Kv acquisition and monoenergetic images ranging from 40 to 190 keV in increments of 10). The left ventricle myocardium was segmented using semiautomated software (Syngo.Via FRONTIER, version 5.0.2; Siemens). Radiomic features were extracted from multiple spectral reconstructions (batches). The consistency of these radiomics features across different batches was evaluated with and without ComBat harmonization using Cohen's d and Principal component analysis (PCA). Both parametric and nonparametric ComBat methods were considered.

RESULTS

Without any ComBat technique, 43.40% of features remained consistent across all multienergy reconstructions. Applying ComBat harmonization increased this consistency to 98.37% with parametric empirical bayes (EB) ComBat and EB M-ComBat, and to 91.52% and 92.33% with nonparametric EB ComBat and nonparametric EB M-ComBat, respectively. PCA without ComBat revealed noticeable differences in the first two principal components between batches, indicating a batch effect or unstable radiomic features. Following ComBat harmonization, the principal components showed more consistency between batches, demonstrating radiomics feature stability between batches.

CONCLUSION

ComBat harmonization enhanced the consistency of radiomic features from multi-energy CT data. Integrating ComBat harmonization may lead to more reproducible results in multienergy CT radiomics studies.

Identifying etiologies of heart failure using non-contrast cardiac magnetic resonance imaging: cine imaging, T1 and T2 mapping, and texture analysis for T1 mapping

Objective

The aim of this retrospective study was to evaluate the usefulness of non-contrast cardiac magnetic resonance imaging, including cine imaging, T1 and T2 mapping, and texture analysis for T1 mapping, for identifying etiologies of heart failure (HF).

Methods

Forty-seven patients with HF were examined using a 1.5 T scanner. Cine imaging parameters and native T1 and T2 values at the mid-septal segment were measured. Vertical run length nonuniformity, vertical gray level nonuniformity (vGLNU), wavelet energy LL(3) and HH (4) on T1 mapping were estimated at the mid-septal segment using open-access software. Late gadolinium enhancement was investigated to help diagnose the etiologies of HF. We used Kruscal-Wallis' with a post-hoc Steel-Dwass' test, Wilcoxon signed-ranked test, Pearson's chai square test and receiver operator curve analysis (ROC) to assess the usefulness of non-contrast CMR for identifying etiologies of HF.

Results

There were significant differences in left ventricular end-diastolic volume (LVEDV) indexed to body surface area (LVEDVi), left ventricular myocardial mass/LVEDV, native T1, and vGLNU between dilated cardiomyopathy (DCM), hypertensive cardiomyopathy (HC) and tachycardia-induced cardiomyopathies (TIC). DCM had higher T1 and lower vGLNU than HC. When compared with TIC, DCM showed significantly higher LVEDV and LVEDVi. ROC analysis revealed that LVEDV and vGLNU provided high specificity for differentiating DCM from the other etiologies.

Conclusion

Native T1 mapping and its texture analysis may be valuable for differentiating between DCM and HC. Cine imaging can be useful for differentiating between DCM and TIC.

Identification of patients with unstable angina based on coronary CT angiography: the application of pericoronary adipose tissue radiomics

Objective

To explore whether radiomics analysis of pericoronary adipose tissue (PCAT) captured by coronary computed tomography angiography (CCTA) could discriminate unstable angina (UA) from stable angina (SA).

Methods

In this single-center retrospective case-control study, coronary CT images and clinical data from 240 angina patients were collected and analyzed. Patients with unstable angina (n = 120) were well-matched with those having stable angina (n = 120). All patients were randomly divided into training (70%) and testing (30%) datasets. Automatic segmentation was performed on the pericoronary adipose tissue surrounding the proximal segments of the left anterior descending artery (LAD), left circumflex coronary artery (LCX), and right coronary artery (RCA). Corresponding radiomic features were extracted and selected, and the fat attenuation index (FAI) for these three vessels was quantified. Machine learning techniques were employed to construct the FAI and radiomic models. Multivariate logistic regression analysis was used to identify the most relevant clinical features, which were then combined with radiomic features to create clinical and integrated models. The performance of different models was compared in terms of area under the curve (AUC), calibration, clinical utility, and sensitivity.

Results

In both training and validation cohorts, the integrated model (AUC = 0.87, 0.74) demonstrated superior discriminatory ability compared to the FAI model (AUC = 0.68, 0.51), clinical feature model (AUC = 0.84, 0.67), and radiomic model (AUC = 0.85, 0.73). The nomogram derived from the combined radiomic and clinical features exhibited excellent performance in diagnosing and predicting unstable angina. Calibration curves showed good fit for all four machine learning models. Decision curve analysis indicated that the integrated model provided better clinical benefit than the other three models.

Conclusions

CCTA-based radiomics signature of PCAT is better than the FAI model in identifying unstable angina and stable angina. The integrated model constructed by combining radiomics and clinical features could further improve the diagnosis and differentiation ability of unstable angina.

CMR to characterize myocardial structure and function in heart failure with preserved left ventricular ejection fraction

Despite remarkable progress in therapeutic drugs, morbidity, and mortality for heart failure (HF) remains high in developed countries. HF with preserved ejection fraction (HFpEF) now accounts for around half of all HF cases. It is a heterogeneous disease, with multiple aetiologies, and as such poses a significant diagnostic challenge. Cardiac magnetic resonance (CMR) has become a valuable non-invasive modality to assess cardiac morphology and function, but beyond that, the multi-parametric nature of CMR allows novel approaches to characterize haemodynamics and with magnetic resonance spectroscopy (MRS), the study of metabolism. Furthermore, exercise CMR, when combined with lung water imaging provides an in-depth understanding of the underlying pathophysiological and mechanistic processes in HFpEF. Thus, CMR provides a comprehensive phenotyping tool for HFpEF, which points towards a targeted and personalized therapy with improved diagnostics and prevention.

Predicting Left Ventricular Adverse Remodeling After Transcatheter Aortic Valve Replacement: A Radiomics Approach

RATIONALE AND OBJECTIVES

To develop a radiomics model based on cardiac computed tomography (CT) for predicting left ventricular adverse remodeling (LVAR) in patients with severe aortic stenosis (AS) who underwent transcatheter aortic valve replacement (TAVR).

MATERIALS AND METHODS

Patients with severe AS who underwent TAVR from January 2019 to December 2022 were recruited. The cohort was divided into adverse remodeling group and non-adverse remodeling group based on LVAR occurrence, and further randomly divided into a training set and a validation set at an 8:2 ratio. Left ventricular radiomics features were extracted from cardiac CT. The least absolute shrinkage and selection operator regression was utilized to select the most relevant radiomics features and clinical features. The radiomics features were used to construct the Radscore, which was then combined with the selected clinical features to build a nomogram. The predictive performance of the models was evaluated using the area under the curve (AUC), while the clinical value of the models was assessed using calibration curves and decision curve analysis.

RESULTS

A total of 273 patients were finally enrolled, including 71 with adverse remodeling and 202 with non-adverse remodeling. 12 radiomics features and five clinical features were extracted to construct the radiomics model, clinical model, and nomogram, respectively. The radiomics model outperformed the clinical model (training AUC: 0.799 vs. 0.760; validation AUC: 0.766 vs. 0.755). The nomogram showed highest accuracy (training AUC: 0.859, validation AUC: 0.837) and was deemed most clinically valuable by decision curve analysis.

CONCLUSION

The cardiac CT-based radiomics features could predict LVAR after TAVR in patients with severe AS.

Genetics of Cardiac Aging Implicate Organ-Specific Variation

Heart structure and function change with age, and the notion that the heart may age faster for some individuals than for others has driven interest in estimating cardiac age acceleration. However, current approaches have limited feature richness (heart measurements; radiomics) or capture extraneous data and therefore lack cardiac specificity (deep learning [DL] on unmasked chest MRI). These technical limitations have been a barrier to efforts to understand genetic contributions to age acceleration. We hypothesized that a video-based DL model provided with heart-masked MRI data would capture a rich yet cardiac-specific representation of cardiac aging. In 61,691 UK Biobank participants, we excluded noncardiac pixels from cardiac MRI and trained a video-based DL model to predict age from one cardiac cycle in the 4-chamber view. We then computed cardiac age acceleration as the bias-corrected prediction of heart age minus the calendar age. Predicted heart age explained 71.1% of variance in calendar age, with a mean absolute error of 3.3 years. Cardiac age acceleration was linked to unfavorable cardiac geometry and systolic and diastolic dysfunction. We also observed links between cardiac age acceleration and diet, decreased physical activity, increased alcohol and tobacco use, and altered levels of 239 serum proteins, as well as adverse brain MRI characteristics. We found cardiac age acceleration to be heritable (h2g 26.6%); a genome-wide association study identified 8 loci related to linked to cardiomyopathy (near TTN, TNS1, LSM3, PALLD, DSP, PLEC, ANKRD1 and MYO18B) and an additional 16 loci (near MECOM, NPR3, KLHL3, HDGFL1, CDKN1A, ELN, SLC25A37, PI15, AP3M1, HMGA2, ADPRHL1, PGAP3, WNT9B, UHRF1 and DOK5). Of the discovered loci, 21 were not previously associated with cardiac age acceleration. Mendelian randomization revealed that lower genetically mediated levels of 6 circulating proteins (MSRA most strongly), as well as greater levels of 5 proteins (LXN most strongly) were associated with cardiac age acceleration, as were greater blood pressure and Lp(a). A polygenic score for cardiac age acceleration predicted earlier onset of arrhythmia, heart failure, myocardial infarction, and mortality. These findings provide a thematic understanding of cardiac age acceleration and suggest that heart- and vascular-specific factors are key to cardiac age acceleration, predominating over a more global aging program.

Improving the efficiency and accuracy of cardiovascular magnetic resonance with artificial intelligence-review of evidence and proposition of a roadmap to clinical translation

BACKGROUND

Cardiovascular magnetic resonance (CMR) is an important imaging modality for the assessment of heart disease; however, limitations of CMR include long exam times and high complexity compared to other cardiac imaging modalities. Recently advancements in artificial intelligence (AI) technology have shown great potential to address many CMR limitations. While the developments are remarkable, translation of AI-based methods into real-world CMR clinical practice remains at a nascent stage and much work lies ahead to realize the full potential of AI for CMR.

METHODS

Herein we review recent cutting-edge and representative examples demonstrating how AI can advance CMR in areas such as exam planning, accelerated image reconstruction, post-processing, quality control, classification and diagnosis.

RESULTS

These advances can be applied to speed up and simplify essentially every application including cine, strain, late gadolinium enhancement, parametric mapping, 3D whole heart, flow, perfusion and others. AI is a unique technology based on training models using data. Beyond reviewing the literature, this paper discusses important AI-specific issues in the context of CMR, including (1) properties and characteristics of datasets for training and validation, (2) previously published guidelines for reporting CMR AI research, (3) considerations around clinical deployment, (4) responsibilities of clinicians and the need for multi-disciplinary teams in the development and deployment of AI in CMR, (5) industry considerations, and (6) regulatory perspectives.

CONCLUSIONS

Understanding and consideration of all these factors will contribute to the effective and ethical deployment of AI to improve clinical CMR.

Perfect Match: Radiomics and Artificial Intelligence in Cardiac Imaging

Cardiovascular diseases remain a significant health burden, with imaging modalities like echocardiography, cardiac computed tomography, and cardiac magnetic resonance imaging playing a crucial role in diagnosis and prognosis. However, the inherent heterogeneity of these diseases poses challenges, necessitating advanced analytical methods like radiomics and artificial intelligence. Radiomics extracts quantitative features from medical images, capturing intricate patterns and subtle variations that may elude visual inspection. Artificial intelligence techniques, including deep learning, can analyze these features to generate knowledge, define novel imaging biomarkers, and support diagnostic decision-making and outcome prediction. Radiomics and artificial intelligence thus hold promise for significantly enhancing diagnostic and prognostic capabilities in cardiac imaging, paving the way for more personalized and effective patient care. This review explores the synergies between radiomics and artificial intelligence in cardiac imaging, following the radiomics workflow and introducing concepts from both domains. Potential clinical applications, challenges, and limitations are discussed, along with solutions to overcome them.

A study on the application of radiomics based on cardiac MR non-enhanced cine sequence in the early diagnosis of hypertensive heart disease

BACKGROUND

The prevalence of hypertensive heart disease (HHD) is high and there is currently no easy way to detect early HHD. Explore the application of radiomics using cardiac magnetic resonance (CMR) non-enhanced cine sequences in diagnosing HHD and latent cardiac changes caused by hypertension.

METHODS

132 patients who underwent CMR scanning were divided into groups: HHD (42), hypertension with normal cardiac structure and function (HWN) group (46), and normal control (NOR) group (44). Myocardial regions of the end-diastolic (ED) and end-systolic (ES) phases of the CMR short-axis cine sequence images were segmented into regions of interest (ROI). Three feature subsets (ED, ES, and ED combined with ES) were established after radiomic least absolute shrinkage and selection operator feature selection. Nine radiomic models were built using random forest (RF), support vector machine (SVM), and naive Bayes. Model performance was analyzed using receiver operating characteristic curves, and metrics like accuracy, area under the curve (AUC), precision, recall, and specificity.

RESULTS

The feature subsets included first-order, shape, and texture features. SVM of ED combined with ES achieved the highest accuracy (0.833), with a macro-average AUC of 0.941. AUCs for HHD, HWN, and NOR identification were 0.967, 0.876, and 0.963, respectively. Precisions were 0.972, 0.740, and 0.826; recalls were 0.833, 0.804, and 0.863, respectively; and specificities were 0.989, 0.863, and 0.909, respectively.

CONCLUSIONS

Radiomics technology using CMR non-enhanced cine sequences can detect early cardiac changes due to hypertension. It holds promise for future use in screening for latent cardiac damage in early HHD.

Pericoronary radiomics texture features associated with hypercholesterolemia on a photon-counting-CT

Introduction

Pericoronary adipose tissue (PCAT) stands in complex bidirectional interaction with the surrounding arteries and is known to be connected to many cardiovascular diseases involving vascular inflammation. PCAT texture may be influenced by other cardiovascular risk factors such as hypercholesterolemia. The recently established photon-counting CT could improve texture analysis and help detect those changes by offering higher spatial resolution and signal-to-noise ratio.

Methods

In this retrospective, single-center, IRB-approved study, PCAT of the left and right coronary artery was manually segmented and radiomic features were extracted using pyradiomics. The study population consisted of a test collective and a validation collective. The collectives were each divided into two groups defined by the presence or absence of hypercholesterolemia, taken from self-reported conditions and confirmed by medical records. Mean and standard deviation were calculated with Pearson correlation coefficient for correlation of features and visualized as boxplots and heatmaps using R statistics. Random forest feature selection was performed to identify differentiating features between the two groups. 66 patients were enrolled in this study (34 female, mean age 58 years).

Results

Two radiomics features allowing differentiation between PCAT texture of the groups were identified (p-values between 0.013 and 0.24) and validated. Patients with hypercholesterolemia presented with a greater concentration of high-density values as indicated through analysis of specific texture features as "gldm_HighGrayLevelEmphasis" (23.95 vs. 22.99) and "glrlm_HighGrayLevelRunEmphasis" (24.21 vs. 23.31).

Discussion

Texture analysis of PCAT allowed differentiation between patients with and without hypercholesterolemia offering a potential imaging biomarker for this specific cardiovascular risk factor.

Prognostic prediction of left ventricular myocardial noncompaction using machine learning and cardiac magnetic resonance radiomics

Background

Although there are many studies on the prognostic factors of left ventricular myocardial noncompaction (LVNC), the determinants are varied and not entirely consistent. This study aimed to build predictive models using radiomics features and machine learning to predict major adverse cardiovascular events (MACEs) in patients with LVNC.

Methods

In total, 96 patients with LVNC were included and randomly divided into training and test cohorts. A total of 105 cine cardiac magnetic resonance (CMR)-derived radiomics features and 35 clinical characteristics were extracted. Five different oversampling algorithms were compared for selection of the optimal imbalanced processing. Feature importance was assessed with extreme gradient boosting (XGBoost). We compared the performance of 5 machine learning classification methods with different sample:feature ratios to determine the optimal hybrid classification strategy. Subsequently, radiomics, clinical, and combined radiomics-clinical models were developed and compared.

Results

The machine learning pipeline included an adaptive synthetic (ADASYN) algorithm for imbalanced processing, XGBoost feature selection with a sample:feature ratio of 10, and support vector machine (SVM) modeling. The areas under the receiver operating characteristic curves (AUCs) of the radiomics model, clinical model, and combined model in the validation cohort were 0.87 (sensitivity 83.33%, specificity 64.29%), 0.65 (sensitivity 16.67%, specificity 78.57%), and 0.92 (specificity 33.33%, sensitivity 100.00%), respectively. The radiomics model performed similarly to the clinical and combined models (P=0.124 and P=0.621, respectively). The performance of the combined model was significantly better than that of the clinical model (P=0.003).

Conclusions

The machine learning-based cine CMR radiomics model performed well at predicting MACEs in patients with LVNC. Adding radiomics features offered incremental prognostic value over clinical factors alone.

Phenotyping heart failure by cardiac magnetic resonance imaging of cardiac macro- and microscopic structure: state of the art review

Heart failure demographics have evolved in past decades with the development of improved diagnostics, therapies, and prevention. Cardiac magnetic resonance (CMR) has developed in a similar timeframe to become the gold-standard non-invasive imaging modality for characterizing diseases causing heart failure. CMR techniques to assess cardiac morphology and function have progressed since their first use in the 1980s. Increasingly efficient acquisition protocols generate high spatial and temporal resolution images in less time. This has enabled new methods of characterizing cardiac systolic and diastolic function such as strain analysis, exercise real-time cine imaging and four-dimensional flow. A key strength of CMR is its ability to non-invasively interrogate the myocardial tissue composition. Gadolinium contrast agents revolutionized non-invasive cardiac imaging with the late gadolinium enhancement technique. Further advances enabled quantitative parametric mapping to increase sensitivity at detecting diffuse pathology. Novel methods such as diffusion tensor imaging and artificial intelligence-enhanced image generation are on the horizon. Magnetic resonance spectroscopy (MRS) provides a window into the molecular environment of the myocardium. Phosphorus (31P) spectroscopy can inform the status of cardiac energetics in health and disease. Proton (1H) spectroscopy complements this by measuring creatine and intramyocardial lipids. Hyperpolarized carbon (13C) spectroscopy is a novel method that could further our understanding of dynamic cardiac metabolism. CMR of other organs such as the lungs may add further depth into phenotypes of heart failure. The vast capabilities of CMR should be deployed and interpreted in context of current heart failure challenges.

Cine MRI-Derived Radiomics Features of the Cardiac Blood Pool: Periodicity, Specificity, and Reproducibility

BACKGROUND

Although radiomics features of the left ventricular wall have been used to assess cardiac diseases, radiomics features of the cardiac blood pool have been relatively ignored.

PURPOSE

To test the hypothesis that cine MRI-derived radiomics features of the cardiac blood pool are associated with cardiac function and motion.

STUDY TYPE

Retrospective.

POPULATION

A total of 26 healthy volunteers (51.2 ± 15.6 years; 17 males).

FIELD STRENGTH/SEQUENCE

A 1.5 T/balanced steady-state free precession (bSSFP).

ASSESSMENT

The radiomics features (107 features in seven classes) of the blood pool of the left/right ventricle/atrium (LV/RV/LA/RA) were extracted on four-chamber cine images (25 phases). Conventional cardiac function parameters (volumes, ejection fraction [EF] and longitudinal strain) were assessed in each cardiac chamber. Intraobserver- and interobserver agreements of radiomics features of all chambers acquired at all phases were assessed, as well as scan-rescan agreement in a subset of 13 volunteers.

STATISTICAL TESTS

Pearson correlation coefficients (r) were used to assess the associations between peak values of radiomics features and end-diastolic (or maximal) volume, end-systolic (or minimal) volume, EF, and longitudinal strain of corresponding chambers. Good intraobserver, interobserver, and scan-rescan agreements for radiomics features acquired were defined as intraclass correlation coefficient (ICC) > 0.7 or coefficient of variation (CoV) < 20%.

RESULTS

Most radiomics features of the blood pool varied periodically throughout the cardiac cycle. Peak values of chamber-specific blood pool radiomics features were correlated with traditional cardiac function and motion indices of corresponding chambers (r: 0.4-0.87). Ninety-three (87%), 86 (80%), and 73 (68%) radiomics features demonstrated good intraobserver, interobserver, and scan-rescan reproducibility, respectively.

CONCLUSION

Cine MRI-derived radiomics features within LV/RV/LA/RA are associated with traditional cardiac function and motion indices of corresponding chambers and may have the potential to become novel quantitative imaging biomarkers in cardiovascular medicine.

EVIDENCE LEVEL

3.

TECHNICAL EFFICACY

1.

Radiomics for the detection of diffusely impaired myocardial perfusion: A proof-of-concept study using 13N-ammonia positron emission tomography

AIM

The current proof-of-concept study investigates the value of radiomic features from normal 13N-ammonia positron emission tomography (PET) myocardial retention images to identify patients with reduced global myocardial flow reserve (MFR).

METHODS

Data from 100 patients with normal retention 13N-ammonia PET scans were divided into two groups, according to global MFR (i.e., < 2 and ≥ 2), as derived from quantitative PET analysis. We extracted radiomic features from retention images at each of five different gray-level (GL) discretization (8, 16, 32, 64, and 128 bins). Outcome independent and dependent feature selection and subsequent univariate and multivariate analyses was performed to identify image features predicting reduced global MFR.

RESULTS

A total of 475 radiomic features were extracted per patient. Outcome independent and dependent feature selection resulted in a remainder of 35 features. Discretization at 16 bins (GL16) yielded the highest number of significant predictors of reduced MFR and was chosen for the final analysis. GLRLM_GLNU was the most robust parameter and at a cut-off of 948 yielded an accuracy, sensitivity, specificity, negative and positive predictive value of 67%, 74%, 58%, 64%, and 69%, respectively, to detect diffusely impaired myocardial perfusion.

CONCLUSION

A single radiomic feature (GLRLM_GLNU) extracted from visually normal 13N-ammonia PET retention images independently predicts reduced global MFR with moderate accuracy. This concept could potentially be applied to other myocardial perfusion imaging modalities based purely on relative distribution patterns to allow for better detection of diffuse disease.

Radiomics applications in cardiac imaging: a comprehensive review

Radiomics is a new emerging field that includes extraction of metrics and quantification of so-called radiomic features from medical images. The growing importance of radiomics applied to oncology in improving diagnosis, cancer staging and grading, and improved personalized treatment, has been well established; yet, this new analysis technique has still few applications in cardiovascular imaging. Several studies have shown promising results describing how radiomics principles could improve the diagnostic accuracy of coronary computed tomography angiography (CCTA) and magnetic resonance imaging (MRI) in diagnosis, risk stratification, and follow-up of patients with coronary heart disease (CAD), ischemic heart disease (IHD), hypertrophic cardiomyopathy (HCM), hypertensive heart disease (HHD), and many other cardiovascular diseases. Such quantitative approach could be useful to overcome the main limitations of CCTA and MRI in the evaluation of cardiovascular diseases, such as readers' subjectiveness and lack of repeatability. Moreover, this new discipline could potentially overcome some technical problems, namely the need of contrast administration or invasive examinations. Despite such advantages, radiomics is still not applied in clinical routine, due to lack of standardized parameters acquisition, inconsistent radiomic methods, lack of external validation, and different knowledge and experience among the readers. The purpose of this manuscript is to provide a recent update on the status of radiomics clinical applications in cardiovascular imaging.

Prediction of incident cardiovascular events using machine learning and CMR radiomics

OBJECTIVES

Evaluation of the feasibility of using cardiovascular magnetic resonance (CMR) radiomics in the prediction of incident atrial fibrillation (AF), heart failure (HF), myocardial infarction (MI), and stroke using machine learning techniques.

METHODS

We identified participants from the UK Biobank who experienced incident AF, HF, MI, or stroke during the continuous longitudinal follow-up. The CMR indices and the vascular risk factors (VRFs) as well as the CMR images were obtained for each participant. Three-segmented regions of interest (ROIs) were computed: right ventricle cavity, left ventricle (LV) cavity, and LV myocardium in end-systole and end-diastole phases. Radiomics features were extracted from the 3D volumes of the ROIs. Seven integrative models were built for each incident cardiovascular disease (CVD) as an outcome. Each model was built with VRF, CMR indices, and radiomics features and a combination of them. Support vector machine was used for classification. To assess the model performance, the accuracy, sensitivity, specificity, and AUC were reported.

RESULTS

AF prediction model using the VRF+CMR+Rad model (accuracy: 0.71, AUC 0.76) obtained the best result. However, the AUC was similar to the VRF+Rad model. HF showed the most significant improvement with the inclusion of CMR metrics (VRF+CMR+Rad: 0.79, AUC 0.84). Moreover, adding only the radiomics features to the VRF reached an almost similarly good performance (VRF+Rad: accuracy 0.77, AUC 0.83). Prediction models looking into incident MI and stroke reached slightly smaller improvement.

CONCLUSIONS

Radiomics features may provide incremental predictive value over VRF and CMR indices in the prediction of incident CVDs.

KEY POINTS

• Prediction of incident atrial fibrillation, heart failure, stroke, and myocardial infarction using machine learning techniques. • CMR radiomics, vascular risk factors, and standard CMR indices will be considered in the machine learning models. • The experiments show that radiomics features can provide incremental predictive value over VRF and CMR indices in the prediction of incident cardiovascular diseases.

Attri-VAE: Attribute-based interpretable representations of medical images with variational autoencoders

Deep learning (DL) methods where interpretability is intrinsically considered as part of the model are required to better understand the relationship of clinical and imaging-based attributes with DL outcomes, thus facilitating their use in the reasoning behind the medical decisions. Latent space representations built with variational autoencoders (VAE) do not ensure individual control of data attributes. Attribute-based methods enforcing attribute disentanglement have been proposed in the literature for classical computer vision tasks in benchmark data. In this paper, we propose a VAE approach, the Attri-VAE, that includes an attribute regularization term to associate clinical and medical imaging attributes with different regularized dimensions in the generated latent space, enabling a better-disentangled interpretation of the attributes. Furthermore, the generated attention maps explained the attribute encoding in the regularized latent space dimensions. Using the Attri-VAE approach we analyzed healthy and myocardial infarction patients with clinical, cardiac morphology, and radiomics attributes. The proposed model provided an excellent trade-off between reconstruction fidelity, disentanglement, and interpretability, outperforming state-of-the-art VAE approaches according to several quantitative metrics. The resulting latent space allowed the generation of realistic synthetic data in the trajectory between two distinct input samples or along a specific attribute dimension to better interpret changes between different cardiac conditions.

Machine Learning Based on Computed Tomography Pulmonary Angiography in Evaluating Pulmonary Artery Pressure in Patients with Pulmonary Hypertension

BACKGROUND

Right heart catheterization is the gold standard for evaluating hemodynamic parameters of pulmonary circulation, especially pulmonary artery pressure (PAP) for diagnosis of pulmonary hypertension (PH). However, the invasive and costly nature of RHC limits its widespread application in daily practice.

PURPOSE

To develop a fully automatic framework for PAP assessment via machine learning based on computed tomography pulmonary angiography (CTPA).

MATERIALS AND METHODS

A machine learning model was developed to automatically extract morphological features of pulmonary artery and the heart on CTPA cases collected between June 2017 and July 2021 based on a single center experience. Patients with PH received CTPA and RHC examinations within 1 week. The eight substructures of pulmonary artery and heart were automatically segmented through our proposed segmentation framework. Eighty percent of patients were used for the training data set and twenty percent for the independent testing data set. PAP parameters, including mPAP, sPAP, dPAP, and TPR, were defined as ground-truth. A regression model was built to predict PAP parameters and a classification model to separate patients through mPAP and sPAP with cut-off values of 40 mm Hg and 55 mm Hg in PH patients, respectively. The performances of the regression model and the classification model were evaluated by analyzing the intraclass correlation coefficient (ICC) and the area under the receiver operating characteristic curve (AUC).

RESULTS

Study participants included 55 patients with PH (men 13; age 47.75 ± 14.87 years). The average dice score for segmentation increased from 87.3% ± 2.9 to 88.2% ± 2.9 through proposed segmentation framework. After features extraction, some of the AI automatic extractions (AAd, RVd, LAd, and RPAd) achieved good consistency with the manual measurements. The differences between them were not statistically significant (t = 1.222, p = 0.227; t = -0.347, p = 0.730; t = 0.484, p = 0.630; t = -0.320, p = 0.750, respectively). The Spearman test was used to find key features which are highly correlated with PAP parameters. Correlations between pulmonary artery pressure and CTPA features show a high correlation between mPAP and LAd, LVd, LAa (r = 0.333, p = 0.012; r = -0.400, p = 0.002; r = -0.208, p = 0.123; r = -0.470, p = 0.000; respectively). The ICC between the output of the regression model and the ground-truth from RHC of mPAP, sPAP, and dPAP were 0.934, 0.903, and 0.981, respectively. The AUC of the receiver operating characteristic curve of the classification model of mPAP and sPAP were 0.911 and 0.833.

CONCLUSIONS

The proposed machine learning framework on CTPA enables accurate segmentation of pulmonary artery and heart and automatic assessment of the PAP parameters and has the ability to accurately distinguish different PH patients with mPAP and sPAP. Results of this study may provide additional risk stratification indicators in the future with non-invasive CTPA data.

UK Biobank: a globally important resource for cancer research

UK Biobank is a large-scale prospective study with deep phenotyping and genomic data. Its open-access policy allows researchers worldwide, from academia or industry, to perform health research in the public interest. Between 2006 and 2010, the study recruited 502,000 adults aged 40-69 years from the general population of the United Kingdom. At enrolment, participants provided information on a wide range of factors, physical measurements were taken, and biological samples (blood, urine and saliva) were collected for long-term storage. Participants have now been followed up for over a decade with more than 52,000 incident cancer cases recorded. The study continues to be enhanced with repeat assessments, web-based questionnaires, multi-modal imaging, and conversion of the stored biological samples to genomic and other '-omic' data. The study has already demonstrated its value in enabling research into the determinants of cancer, and future planned enhancements will make the resource even more valuable to cancer researchers. Over 26,000 researchers worldwide are currently using the data, performing a wide range of cancer research. UK Biobank is uniquely placed to transform our understanding of the causes of cancer development and progression, and drive improvements in cancer treatment and prevention over the coming decades.

Evaluation of radiomics feature stability in abdominal monoenergetic photon counting CT reconstructions

Feature stability and standardization remain challenges that impede the clinical implementation of radiomics. This study investigates the potential of spectral reconstructions from photon-counting computed tomography (PCCT) regarding organ-specific radiomics feature stability. Abdominal portal-venous phase PCCT scans of 10 patients in virtual monoenergetic (VM) (keV 40-120 in steps of 10), polyenergetic, virtual non-contrast (VNC), and iodine maps were acquired. Two 2D and 3D segmentations measuring 1 and 2 cm in diameter of the liver, lung, spleen, psoas muscle, subcutaneous fat, and air were obtained for spectral reconstructions. Radiomics features were extracted with pyradiomics. The calculation of feature-specific intraclass correlation coefficients (ICC) was performed by comparing all segmentation approaches and organs. Feature-wise and organ-wise correlations were evaluated. Segmentation-resegmentation stability was evaluated by concordance correlation coefficient (CCC). Compared to non-VM, VM-reconstruction features tended to be more stable. For VM reconstructions, 3D 2 cm segmentation showed the highest average ICC with 0.63. Based on a criterion of ≥ 3 stable organs and an ICC of ≥ 0.75, 12-mainly non-first-order features-are shown to be stable between the VM reconstructions. In a segmentation-resegmentation analysis in 3D 2 cm, three features were identified as stable based on a CCC of > 0.6 in ≥ 3 organs in ≥ 6 VM reconstructions. Certain radiomics features vary between monoenergetic reconstructions and depend on the ROI size. Feature stability was also shown to differ between different organs. Yet, glcm_JointEntropy, gldm_GrayLevelNonUniformity, and firstorder_Entropy could be identified as features that could be interpreted as energy-independent and segmentation-resegmentation stable in this PCCT collective. PCCT may support radiomics feature standardization and comparability between sites.

Differentiating between cardiac amyloidosis and hypertrophic cardiomyopathy on non-contrast cine-magnetic resonance images using machine learning-based radiomics

Objectives

This study aimed to determine whether texture analysis (TA) and machine learning-based classifications can be applied in differential diagnosis of cardiac amyloidosis (CA) and hypertrophic cardiomyopathy (HCM) using non-contrast cine cardiac magnetic resonance (CMR) images.

Methods

In this institutional review board-approved study, we consecutively enrolled 167 patients with CA (n = 85), HCM (n = 82), and 84 patients with normal CMR served as controls. All cases were randomized into training [119 patients (70%)] and validation [48 patients (30%)] groups. A total of 275 texture features were extracted from cine images. Based on regression analysis with the least absolute shrinkage and selection operator (LASSO), nine machine learning models were established and their diagnostic performance determined.

Results

Nineteen radiomics texture features derived from cine images were used to differentiate CA and HCM. In the validation cohort, the support vector machine (SVM), which had an accuracy of 0.85, showed the best performance (MCC = 0.637). Gray level non-uniformity (GLevNonU) was the single most effective feature. The combined model of radiomics texture features and conventional MR metrics had superior discriminatory performance (AUC = 0.89) over conventional MR metrics model (AUC = 0.79). Moreover, results showed that GLevNonU levels in HCM patients were significantly higher compared with levels in CA patients and control groups (P < 0.001). A cut-off of GLevNonU ≥ 25 was shown to differentiate between CA and HCM patients, with an area under the curve (AUC) of 0.86 (CI:0.804–0.920). Multiple comparisons tests showed that GLevNonU was significantly greater in LGE+, relative to LGE-patient groups (CA+ vs. CA- and HCM+ vs. HCM-, P = 0.01, 0.001, respectively).

Conclusion

Machine learning-based classifiers can accurately differentiate between CA and HCM on non-contrast cine images. The radiomics-MR combined model can be used to improve the discriminatory performance. TA may be used to assess myocardial microstructure changes that occur during different stages of cardiomyopathies.

Evaluation of MRI-based radiomic features in heart morphologic variations as a consequence of autoimmune thyroid disorders

Radiomics (RC) was initially developed using computed tomography (CT) for oncological imaging. However, it can be applied to various scientific and clinical radiology fields regardless of the modalities involved. The purpose of this survey was to evaluate alterations in magnetic resonance imaging of the heart (CMR) in patients suffering from autoimmune thyroid disorders (AITD) by applying RC tools and mapping features. A total of 50 individuals were evaluated in this study. We searched for CMR examinations performed in our department between January 2019 and 2021 in patients with AITD. Thirty patients with AITD (21 men and 9 women, aged 51 to 78 years; mean age, 60 years) were enrolled in our survey. We enrolled a control group (CG) of 20 individuals (14 men and 6 women aged 53-87 years; mean age, 68 years) without AITD or cardiac disorders. Global native T1 and T2 mapping revealed no significant differences between groups. However, we identified significantly higher values of several texture parameters, including the gray-level co-occurrence matrix (GLCM) entropy, gray-level run-length matrix (GLRLM; short-run high gray-level emphasis (SRHGE), GLCM (Energy), gray-level size zone matrix length matrix (GLZLM; LZLGE), GLZLM (SZLGE), DISCRETIZED (HISTO-Energy) GLCM (Dissimilarity), and GLCM (Contrast), in patients with AITD in comparison to the CG (P < .01). Our results indicate that several RC properties extracted from CMR images can be used to discriminate between the AITD and CG groups.

Myocardial Radiomics Texture Features Associated with Increased Coronary Calcium Score—First Results of a Photon-Counting CT

The coronary artery calcium score is an independent risk factor of the development of adverse cardiac events. The severity of coronary artery calcification may influence the myocardial texture. Due to higher spatial resolution and signal-to-noise ratio, new CT technologies such as PCCT may improve the detection of texture alterations depending on the severity of coronary artery calcification. In this retrospective, single-center, IRB-approved study, left ventricular myocardium was segmented and radiomics features were extracted using pyradiomics. The mean and standard deviation with the Pearson correlation coefficient for correlations of features were calculated and visualized as boxplots and heatmaps. Random forest feature selection was performed. Thirty patients (26.7% women, median age 58 years) were enrolled in the study. Patients were divided into two subgroups depending on the severity of coronary artery calcification (Agatston score 0 and Agatston score ≥ 100). Through random forest feature selection, a set of four higher-order features could be defined to discriminate myocardial texture between the two groups. When including the additional Agatston 1–99 groups as a validation, a severity-associated change in feature intensity was detected. A subset of radiomics features texture alterations of the left ventricular myocardium was associated with the severity of coronary artery calcification estimated by the Agatston score.

A Collaborative Approach for the Development and Application of Machine Learning Solutions for CMR-Based Cardiac Disease Classification

The quality and acceptance of machine learning (ML) approaches in cardiovascular data interpretation depends strongly on model design and training and the interaction with the clinical experts. We hypothesize that a software infrastructure for the training and application of ML models can support the improvement of the model training and provide relevant information for understanding the classification-relevant data features. The presented solution supports an iterative training, evaluation, and exploration of machine-learning-based multimodal data interpretation methods considering cardiac MRI data. Correction, annotation, and exploration of clinical data and interpretation of results are supported through dedicated interactive visual analytics tools. We test the presented concept with two use cases from the ACDC and EMIDEC cardiac MRI image analysis challenges. In both applications, pre-trained 2D U-Nets are used for segmentation, and classifiers are trained for diagnostic tasks using radiomics features of the segmented anatomical structures. The solution was successfully used to identify outliers in automatic segmentation and image acquisition. The targeted curation and addition of expert annotations improved the performance of the machine learning models. Clinical experts were supported in understanding specific anatomical and functional characteristics of the assigned disease classes.

Quality of science and reporting for radiomics in cardiac magnetic resonance imaging studies: a systematic review

OBJECTIVES

To evaluate the quality of radiomics studies using cardiac magnetic resonance imaging (CMR) according to the radiomics quality score (RQS), Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines, and the standards defined by the Image Biomarker Standardization Initiative (IBSI) and identify areas needing improvement.

MATERIALS AND METHODS

PubMed and Embase were searched to identify radiomics studies using CMR until March 10, 2021. Of the 259 identified articles, 32 relevant original research articles were included. Studies were scored according to the RQS, TRIPOD guidelines, and IBSI standards by two cardiac radiologists.

RESULTS

The mean RQS was 14.3% of the maximum (5.16 out of 36). RQS were low for the demonstration of validation (-60.6%), calibration statistics (1.6%), potential clinical utility (3.1%), and open science (3.1%) items. No study conducted a phantom study or cost-effectiveness analysis. The adherence to TRIPOD guidelines was 55.9%. Studies were deficient in reporting title (3.1%), stating objective in abstract and introduction (6.3% and 9.4%), missing data (0%), discrimination/calibration (3.1%), and how to use the prediction model (3.1%). According to the IBSI standards, non-uniformity correction, image interpolation, grey-level discretization, and signal intensity normalization were performed in two (6.3%), four (12.5%), six (18.8%), and twelve (37.5%) studies, respectively.

CONCLUSION

The quality of radiomics studies using CMR is suboptimal. Improvements are needed in the areas of validation, calibration, clinical utility, and open science. Complete reporting of study objectives, missing data, discrimination/calibration, how to use the prediction model, and preprocessing steps are necessary.

KEY POINTS

• The quality of science in radiomics studies using CMR is currently inadequate. • RQS were low for validation, calibration, clinical utility, and open science; no study conducted a phantom study or cost-effectiveness analysis. • In stating the study objective, missing data, discrimination/calibration, how to use the prediction model, and preprocessing steps, improvements are needed.

Radiomics in Cardiovascular Disease Imaging: from Pixels to the Heart of the Problem

Purpose of Review

This review of the literature aims to present potential applications of radiomics in cardiovascular radiology and, in particular, in cardiac imaging.

Recent Findings

Radiomics and machine learning represent a technological innovation which may be used to extract and analyze quantitative features from medical images. They aid in detecting hidden pattern in medical data, possibly leading to new insights in pathophysiology of different medical conditions. In the recent literature, radiomics and machine learning have been investigated for numerous potential applications in cardiovascular imaging. They have been proposed to improve image acquisition and reconstruction, for anatomical structure automated segmentation or automated characterization of cardiologic diseases.

Summary

The number of applications for radiomics and machine learning is continuing to rise, even though methodological and implementation issues still limit their use in daily practice. In the long term, they may have a positive impact in patient management.

Radiomics Analysis Derived From LGE-MRI Predict Sudden Cardiac Death in Participants With Hypertrophic Cardiomyopathy

Objectives: To identify significant radiomics features derived from late gadolinium enhancement (LGE) images in participants with hypertrophic cardiomyopathy (HCM) and assess their prognostic value in predicting sudden cardiac death (SCD) endpoint.

Method: The 157 radiomic features of 379 sequential participants with HCM who underwent cardiovascular magnetic resonance imaging (MRI) were extracted. CoxNet (Least Absolute Shrinkage and Selection Operator (LASSO) and Elastic Net) and Random Forest models were applied to optimize feature selection for the SCD risk prediction and cross-validation was performed.

Results: During a median follow-up of 29 months (interquartile range, 20–42 months), 27 participants with HCM experienced SCD events. Cox analysis revealed that two selected features, local binary patterns (LBP) (19) (hazard ratio (HR), 1.028, 95% CI: 1.032–1.134; P = 0.001) and Moment (1) (HR, 1.212, 95%CI: 1.032–1.423; P = 0.02) provided significant prognostic value to predict the SCD endpoints after adjustment for the clinical risk predictors and late gadolinium enhancement. Furthermore, the univariately significant risk predictor was improved by the addition of the selected radiomics features, LBP (19) and Moment (1), to predict SCD events (P < 0.05).

Conclusion: The radiomics features of LBP (19) and Moment (1) extracted from LGE images, reflecting scar heterogeneity, have independent prognostic value in identifying high SCD risk patients with HCM.

Cardiac Magnetic Resonance Radiomics Reveal Differential Impact of Sex, Age, and Vascular Risk Factors on Cardiac Structure and Myocardial Tissue

Background: Cardiovascular magnetic resonance (CMR) radiomics analysis provides multiple quantifiers of ventricular shape and myocardial texture, which may be used for detailed cardiovascular phenotyping. Objectives: We studied variation in CMR radiomics phenotypes by age and sex in healthy UK Biobank participants. Then, we examined independent associations of classical vascular risk factors (VRFs: smoking, diabetes, hypertension, high cholesterol) with CMR radiomics features, considering potential sex and age differential relationships. Design: Image acquisition was with 1.5 Tesla scanners (MAGNETOM Aera, Siemens). Three regions of interest were segmented from short axis stack images using an automated pipeline: right ventricle, left ventricle, myocardium. We extracted 237 radiomics features from each study using Pyradiomics. In a healthy subset of participants (n = 14,902) without cardiovascular disease or VRFs, we estimated independent associations of age and sex with each radiomics feature using linear regression models adjusted for body size. We then created a sample comprising individuals with at least one VRF matched to an equal number of healthy participants (n = 27,400). We linearly modelled each radiomics feature against age, sex, body size, and all the VRFs. Bonferroni adjustment for multiple testing was applied to all p-values. To aid interpretation, we organised the results into six feature clusters. Results: Amongst the healthy subset, men had larger ventricles with dimmer and less texturally complex myocardium than women. Increasing age was associated with smaller ventricles and greater variation in myocardial intensities. Broadly, all the VRFs were associated with dimmer, less varied signal intensities, greater uniformity of local intensity levels, and greater relative presence of low signal intensity areas within the myocardium. Diabetes and high cholesterol were also associated with smaller ventricular size, this association was of greater magnitude in men than women. The pattern of alteration of radiomics features with the VRFs was broadly consistent in men and women. However, the associations between intensity based radiomics features with both diabetes and hypertension were more prominent in women than men. Conclusions: We demonstrate novel independent associations of sex, age, and major VRFs with CMR radiomics phenotypes. Further studies into the nature and clinical significance of these phenotypes are needed.

Radiogenomics and Artificial Intelligence Approaches Applied to Cardiac Computed Tomography Angiography and Cardiac Magnetic Resonance for Precision Medicine in Coronary Heart Disease: A Systematic Review

The risk of coronary heart disease (CHD) clinical manifestations and patient management is estimated according to risk scores accounting multifactorial risk factors, thus failing to cover the individual cardiovascular risk. Technological improvements in the field of medical imaging, in particular, in cardiac computed tomography angiography and cardiac magnetic resonance protocols, laid the development of radiogenomics. Radiogenomics aims to integrate a huge number of imaging features and molecular profiles to identify optimal radiomic/biomarker signatures. In addition, supervised and unsupervised artificial intelligence algorithms have the potential to combine different layers of data (imaging parameters and features, clinical variables and biomarkers) and elaborate complex and specific CHD risk models allowing more accurate diagnosis and reliable prognosis prediction. Literature from the past 5 years was systematically collected from PubMed and Scopus databases, and 60 studies were selected. We speculated the applicability of radiogenomics and artificial intelligence through the application of machine learning algorithms to identify CHD and characterize atherosclerotic lesions and myocardial abnormalities. Radiomic features extracted by cardiac computed tomography angiography and cardiac magnetic resonance showed good diagnostic accuracy for the identification of coronary plaques and myocardium structure; on the other hand, few studies exploited radiogenomics integration, thus suggesting further research efforts in this field. Cardiac computed tomography angiography resulted the most used noninvasive imaging modality for artificial intelligence applications. Several studies provided high performance for CHD diagnosis, classification, and prognostic assessment even though several efforts are still needed to validate and standardize algorithms for CHD patient routine according to good medical practice.

Multi-Centre, Multi-Vendor and Multi-Disease Cardiac Segmentation: The M&Ms Challenge

The emergence of deep learning has considerably advanced the state-of-the-art in cardiac magnetic resonance (CMR) segmentation. Many techniques have been proposed over the last few years, bringing the accuracy of automated segmentation close to human performance. However, these models have been all too often trained and validated using cardiac imaging samples from single clinical centres or homogeneous imaging protocols. This has prevented the development and validation of models that are generalizable across different clinical centres, imaging conditions or scanner vendors. To promote further research and scientific benchmarking in the field of generalizable deep learning for cardiac segmentation, this paper presents the results of the Multi-Centre, Multi-Vendor and Multi-Disease Cardiac Segmentation (M&Ms) Challenge, which was recently organized as part of the MICCAI 2020 Conference. A total of 14 teams submitted different solutions to the problem, combining various baseline models, data augmentation strategies, and domain adaptation techniques. The obtained results indicate the importance of intensity-driven data augmentation, as well as the need for further research to improve generalizability towards unseen scanner vendors or new imaging protocols. Furthermore, we present a new resource of 375 heterogeneous CMR datasets acquired by using four different scanner vendors in six hospitals and three different countries (Spain, Canada and Germany), which we provide as open-access for the community to enable future research in the field.

Radiomics-Based Classification of Left Ventricular Non-compaction, Hypertrophic Cardiomyopathy, and Dilated Cardiomyopathy in Cardiovascular Magnetic Resonance

Left Ventricular (LV) Non-compaction (LVNC), Hypertrophic Cardiomyopathy (HCM), and Dilated Cardiomyopathy (DCM) share morphological and functional traits that increase the diagnosis complexity. Additional clinical information, besides imaging data such as cardiovascular magnetic resonance (CMR), is usually required to reach a definitive diagnosis, including electrocardiography (ECG), family history, and genetics. Alternatively, indices of hypertrabeculation have been introduced, but they require tedious and time-consuming delineations of the trabeculae on the CMR images. In this paper, we propose a radiomics approach to automatically encode differences in the underlying shape, gray-scale and textural information in the myocardium and its trabeculae, which may enhance the capacity to differentiate between these overlapping conditions. A total of 118 subjects, including 35 patients with LVNC, 25 with HCM, 37 with DCM, as well as 21 healthy volunteers (NOR), underwent CMR imaging. A comprehensive radiomics characterization was applied to LV short-axis images to quantify shape, first-order, co-occurrence matrix, run-length matrix, and local binary patterns. Conventional CMR indices (LV volumes, mass, wall thickness, LV ejection fraction—LVEF—), as well as hypertrabeculation indices by Petersen and Jacquier, were also analyzed. State-of-the-art Machine Learning (ML) models (one-vs.-rest Support Vector Machine—SVM—, Logistic Regression—LR—, and Random Forest Classifier—RF—) were used for one-vs.-rest classification tasks. The use of radiomics models for the automated diagnosis of LVNC, HCM, and DCM resulted in excellent one-vs.-rest ROC-AUC values of 0.95 while generating these results without the need for the delineation of the trabeculae. First-order and texture features resulted to be among the most discriminative features in the obtained radiomics signatures, indicating their added value for quantifying relevant tissue patterns in cardiomyopathy differential diagnosis.

New Imaging Signatures of Cardiac Alterations in Ischaemic Heart Disease and Cerebrovascular Disease Using CMR Radiomics

Background: Ischaemic heart disease (IHD) and cerebrovascular disease are two closely inter-related clinical entities. Cardiovascular magnetic resonance (CMR) radiomics may capture subtle cardiac changes associated with these two diseases providing new insights into the brain-heart interactions. Objective: To define the CMR radiomics signatures for IHD and cerebrovascular disease and study their incremental value for disease discrimination over conventional CMR indices. Methods: We analysed CMR images of UK Biobank's subjects with pre-existing IHD, ischaemic cerebrovascular disease, myocardial infarction (MI), and ischaemic stroke (IS) (n = 779, 267, 525, and 107, respectively). Each disease group was compared with an equal number of healthy controls. We extracted 446 shape, first-order, and texture radiomics features from three regions of interest (right ventricle, left ventricle, and left ventricular myocardium) in end-diastole and end-systole defined from segmentation of short-axis cine images. Systematic feature selection combined with machine learning (ML) algorithms (support vector machine and random forest) and 10-fold cross-validation tests were used to build the radiomics signature for each condition. We compared the discriminatory power achieved by the radiomics signature with conventional indices for each disease group, using the area under the curve (AUC), receiver operating characteristic (ROC) analysis, and paired t-test for statistical significance. A third model combining both radiomics and conventional indices was also evaluated. Results: In all the study groups, radiomics signatures provided a significantly better disease discrimination than conventional indices, as suggested by AUC (IHD:0.82 vs. 0.75; cerebrovascular disease: 0.79 vs. 0.77; MI: 0.87 vs. 0.79, and IS: 0.81 vs. 0.72). Similar results were observed with the combined models. In IHD and MI, LV shape radiomics were dominant. However, in IS and cerebrovascular disease, the combination of shape and intensity-based features improved the disease discrimination. A notable overlap of the radiomics signatures of IHD and cerebrovascular disease was also found. Conclusions: This study demonstrates the potential value of CMR radiomics over conventional indices in detecting subtle cardiac changes associated with chronic ischaemic processes involving the brain and heart, even in the presence of more heterogeneous clinical pictures. Radiomics analysis might also improve our understanding of the complex mechanisms behind the brain-heart interactions during ischaemia.

Emerging Techniques in Cardiac Magnetic Resonance Imaging

Cardiovascular disease is the leading cause of death and a significant contributor of health care costs. Noninvasive imaging plays an essential role in the management of patients with cardiovascular disease. Cardiac magnetic resonance (MR) can noninvasively assess heart and vascular abnormalities, including biventricular structure/function, blood hemodynamics, myocardial tissue composition, microstructure, perfusion, metabolism, coronary microvascular function, and aortic distensibility/stiffness. Its ability to characterize myocardial tissue composition is unique among alternative imaging modalities in cardiovascular disease. Significant growth in cardiac MR utilization, particularly in Europe in the last decade, has laid the necessary clinical groundwork to position cardiac MR as an important imaging modality in the workup of patients with cardiovascular disease. Although lack of availability, limited training, physician hesitation, and reimbursement issues have hampered widespread clinical adoption of cardiac MR in the United States, growing clinical evidence will ultimately overcome these challenges. Advances in cardiac MR techniques, particularly faster image acquisition, quantitative myocardial tissue characterization, and image analysis have been critical to its growth. In this review article, we discuss recent advances in established and emerging cardiac MR techniques that are expected to strengthen its capability in managing patients with cardiovascular disease. LEVEL OF EVIDENCE: 5 TECHNICAL EFFICACY: Stage 1.

Machine learning phenotyping of scarred myocardium from cine in hypertrophic cardiomyopathy

AIMS

Cardiovascular magnetic resonance (CMR) with late-gadolinium enhancement (LGE) is increasingly being used in hypertrophic cardiomyopathy (HCM) for diagnosis, risk stratification, and monitoring. However, recent data demonstrating brain gadolinium deposits have raised safety concerns. We developed and validated a machine-learning (ML) method that incorporates features extracted from cine to identify HCM patients without fibrosis in whom gadolinium can be avoided.

METHODS AND RESULTS

An XGBoost ML model was developed using regional wall thickness and thickening, and radiomic features of myocardial signal intensity, texture, size, and shape from cine. A CMR dataset containing 1099 HCM patients collected using 1.5T CMR scanners from different vendors and centres was used for model development (n=882) and validation (n=217). Among the 2613 radiomic features, we identified 7 features that provided best discrimination between +LGE and -LGE using 10-fold stratified cross-validation in the development cohort. Subsequently, an XGBoost model was developed using these radiomic features, regional wall thickness and thickening. In the independent validation cohort, the ML model yielded an area under the curve of 0.83 (95% CI: 0.77-0.89), sensitivity of 91%, specificity of 62%, F1-score of 77%, true negatives rate (TNR) of 34%, and negative predictive value (NPV) of 89%. Optimization for sensitivity provided sensitivity of 96%, F2-score of 83%, TNR of 19% and NPV of 91%; false negatives halved from 4% to 2%.

CONCLUSION

An ML model incorporating novel radiomic markers of myocardium from cine can rule-out myocardial fibrosis in one-third of HCM patients referred for CMR reducing unnecessary gadolinium administration.