Texture analysis and machine learning of non-contrast T1-weighted MR images in patients with hypertrophic cardiomyopathy-Preliminary results

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.

Texture Analysis of Native T1 Images as a Novel Method for Noninvasive Assessment of Uremic Cardiomyopathy

BACKGROUND

Noncontrast cardiac T₁ times are increased in dialysis patients which might indicate fibrotic alterations in uremic cardiomyopathy.

PURPOSE

To explore the application of the texture analysis (TA) of T₁ images in the assessment of myocardial alterations in dialysis patients.

STUDY TYPE

Case-control study.

POPULATION

A total of 117 subjects, including 22 on hemodialysis, 44 on peritoneal dialysis, and 51 healthy controls.

FIELD STRENGTH

A 3 T, steady-state free precession (SSFP) sequence, modified Look-Locker imaging (MOLLI).

ASSESSMENT

Two independent, blinded researchers manually delineated endocardial and epicardial borders of the left ventricle (LV) on midventricular T₁ maps for TA.

STATISTICAL TESTS

Texture feature selection was performed, incorporating reproducibility verification, machine learning, and collinearity analysis. Multivariate linear regressions were performed to examine the independent associations between the selected texture features and left ventricular function in dialysis patients. Texture features' performance in discrimination was evaluated by sensitivity and specificity. Reproducibility was estimated by the intraclass correlation coefficient (ICC).

RESULTS

Dialysis patients had greater T₁ values than normal (P < 0.05). Five texture features were filtered out through feature selection, and four showed a statistically significant difference between dialysis patients and healthy controls. Among the four features, vertical run-length nonuniformity (VRLN) had the most remarkable difference among the control and dialysis groups (144 ± 40 vs. 257 ± 74, P < 0.05), which overlap was much smaller than Global T₁ times (1268 ± 38 vs. 1308 ± 46 msec, P < 0.05). The VRLN values were notably elevated (cutoff = 170) in dialysis patients, with a specificity of 97% and a sensitivity of 88%, compared with T₁ times (specificity = 76%, sensitivity = 60%). In dialysis patients, VRLN was significantly and independently associated with left ventricular ejection fraction (P < 0.05), global longitudinal strain (P < 0.05), radial strain (P < 0.05), and circumferential strain (P < 0.05); however, T₁ was not.

DATA CONCLUSION

The texture features obtained by TA of T₁ images and VRLN may be a better parameter for assessing myocardial alterations than T₁ times.

LEVEL OF EVIDENCE

4 TECHNICAL EFFICACY: Stage 3.

Texture analysis applied in T1 maps and extracellular volume obtained using cardiac MRI in the diagnosis of hypertrophic cardiomyopathy and hypertensive heart disease compared with normal controls

AIM

To assess the potential of texture analysis (TA) applied in T1 maps and extracellular volume (ECV) obtained using cardiac magnetic resonance (CMR) in the diagnosis of hypertrophic cardiomyopathy (HCM) and hypertensive heart disease (HHD) compared with normal controls (NC). Strain parameters were analysed to compare with final TA models.

MATERIALS AND METHODS

This retrospective study included 66 HCM patients, 39 HHD patients, and 41 NC. Step-wise dimension reduction and feature selection were performed by reproducibility, machine learning, collinearity, and multivariable regression analysis to select the texture features that enable diagnosis of and differentiation between HCM and HHD. Strain parameters were calculated by short-axis and three long-axis cine sequences.

RESULTS

Independent features in T1 maps and ECV analysis allowed for the differentiation between patients (HCM and HHD) and NC. Of the best-calculated model, the areas under the receiver operating curve (AUCs) were as follows: 0.969 for T1 map and 0.964 for ECV. To distinguish HCM from HHD, two independent features were screened out for both T1 and ECV maps. The AUCs were as follows: 0.793 for T1 map and 0.894 for ECV. Radial, circumferential, and longitudinal strain parameters could differentiate patients from NC, but only longitudinal strain parameters was significantly different between HCM and HHD.

CONCLUSIONS

Texture analysis of T1 maps and ECV shows high accuracy in differentiating hypertrophic myocardium from NC, and HCM from HHD. Strain parameters are able to demonstrate the difference between patients and NC, but were less impressive in differentiating HCM and HHD.

Radiomics of Non-Contrast-Enhanced T1 Mapping: Diagnostic and Predictive Performance for Myocardial Injury in Acute ST-Segment-Elevation Myocardial Infarction

OBJECTIVE

To evaluate the feasibility of texture analysis on non-contrast-enhanced T1 maps of cardiac magnetic resonance (CMR) imaging for the diagnosis of myocardial injury in acute myocardial infarction (MI).

MATERIALS AND METHODS

This study included 68 patients (57 males and 11 females; mean age, 55.7 ± 10.5 years) with acute ST-segment-elevation MI who had undergone 3T CMR after a percutaneous coronary intervention. Forty patients of them also underwent a 6-month follow-up CMR. The CMR protocol included T2-weighted imaging, T1 mapping, rest first-pass perfusion, and late gadolinium enhancement. Radiomics features were extracted from the T1 maps using open-source software. Radiomics signatures were constructed with the selected strongest features to evaluate the myocardial injury severity and predict the recovery of left ventricular (LV) longitudinal systolic myocardial contractility.

RESULTS

A total of 1088 segments of the acute CMR images were analyzed; 103 (9.5%) segments showed microvascular obstruction (MVO), and 557 (51.2%) segments showed MI. A total of 640 segments were included in the 6-month follow-up analysis, of which 160 (25.0%) segments showed favorable recovery of LV longitudinal systolic myocardial contractility. Combined radiomics signature and T1 values resulted in a higher diagnostic performance for MVO compared to T1 values alone (area under the curve [AUC] in the training set; 0.88, 0.72, p = 0.031: AUC in the test set; 0.86, 0.71, p002). Combined radiomics signature and T1 values also provided a higher predictive value for LV longitudinal systolic myocardial contractility recovery compared to T1 values (AUC in the training set; 0.76, 0.55, p < 0.001: AUC in the test set; 0.77, 0.60, p < 0.001).

CONCLUSION

The combination of radiomics of non-contrast-enhanced T1 mapping and T1 values could provide higher diagnostic accuracy for MVO. Radiomics also provides incremental value in the prediction of LV longitudinal systolic myocardial contractility at six months.

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

Cardiovascular magnetic resonance (CMR) radiomics is a novel technique for advanced cardiac image phenotyping by analyzing multiple quantifiers of shape and tissue texture. In this paper, we assess, in the largest sample published to date, the performance of CMR radiomics models for identifying changes in cardiac structure and tissue texture due to cardiovascular risk factors. We evaluated five risk factor groups from the first 5,065 UK Biobank participants: hypertension (n = 1,394), diabetes (n = 243), high cholesterol (n = 779), current smoker (n = 320), and previous smoker (n = 1,394). Each group was randomly matched with an equal number of healthy comparators (without known cardiovascular disease or risk factors). Radiomics analysis was applied to short axis images of the left and right ventricles at end-diastole and end-systole, yielding a total of 684 features per study. Sequential forward feature selection in combination with machine learning (ML) algorithms (support vector machine, random forest, and logistic regression) were used to build radiomics signatures for each specific risk group. We evaluated the degree of separation achieved by the identified radiomics signatures using area under curve (AUC), receiver operating characteristic (ROC), and statistical testing. Logistic regression with L1-regularization was the optimal ML model. Compared to conventional imaging indices, radiomics signatures improved the discrimination of risk factor vs. healthy subgroups as assessed by AUC [diabetes: 0.80 vs. 0.70, hypertension: 0.72 vs. 0.69, high cholesterol: 0.71 vs. 0.65, current smoker: 0.68 vs. 0.65, previous smoker: 0.63 vs. 0.60]. Furthermore, we considered clinical interpretation of risk-specific radiomics signatures. For hypertensive individuals and previous smokers, the surface area to volume ratio was smaller in the risk factor vs. healthy subjects; perhaps reflecting a pattern of global concentric hypertrophy in these conditions. In the diabetes subgroup, the most discriminatory radiomics feature was the median intensity of the myocardium at end-systole, which suggests a global alteration at the myocardial tissue level. This study confirms the feasibility and potential of CMR radiomics for deeper image phenotyping of cardiovascular health and disease. We demonstrate such analysis may have utility beyond conventional CMR metrics for improved detection and understanding of the early effects of cardiovascular risk factors on cardiac structure and tissue.

The influence of cardiac motion on radiomics features: radiomics features of non-enhanced CMR cine images greatly vary through the cardiac cycle

OBJECTIVES

The cardiac cycle might impair the reproducibility of radiomics features of cardiac magnetic resonance (CMR) cine images, yet this issue has not been addressed in the previous research. We aim to evaluate whether radiomics features of CMR cine images vary during the cardiac cycle and investigate the reproducibility of radiomics features of CMR cine images.

METHODS

This retrospective study enrolled 59 healthy adults who underwent CMR examination. Two observers segmented the myocardium on a 4D stack of three consecutive mid-ventricular short-axis cine images covering the cardiac cycle. A total of 352 radiomics features were extracted. The coefficient of variation and intraclass correlation coefficient were used to assess the feature variability through the cycle and inter-observer reproducibility, respectively.

RESULTS

Approximately 55% of radiomics features showed large variability through the cardiac cycle. The original features showed more variability than the Laplacian of Gaussian-filtered features (73.8% vs. 48%). The features of 4D stack cine images had a higher proportion of reproducible features (92.0%, 87.7%, and 76.1%) compared with the end-diastolic (77.8%, 62.2%, and 41.7%) and the end-systolic images (81.5%, 74.1%, and 58.8%) for intraclass correlation cut-off values of ³0.80, > 0.85, and > 0.90, respectively.

CONCLUSIONS

Radiomics features of CMR cine images greatly vary during the cardiac cycle. The radiomics features of 4D stack of cine images are more robust compared with end-diastolic and end-systolic cine images in terms of reproducibility. The impact of the cardiac cycle on the reproducibility of the features should be considered when employing CMR cine images radiomics.

KEY POINTS

• There is limited evidence on the impact of cardiac motion on radiomics features of CMR cine images and the reproducibility of the radiomics features of CMR cine images. • Radiomics features of non-enhanced CMR cine images greatly vary during the cardiac cycle, and the number of "reproducible" features shows significant variations according to the cardiac phases. • The impact of cardiac cycle on the reproducibility of the radiomics features should be considered when employing CMR cine images radiomics.

Artificial intelligence in cardiac radiology

Artificial intelligence (AI) is entering the clinical arena, and in the early stage, its implementation will be focused on the automatization tasks, improving diagnostic accuracy and reducing reading time. Many studies investigate the potential role of AI to support cardiac radiologist in their day-to-day tasks, assisting in segmentation, quantification, and reporting tasks. In addition, AI algorithms can be also utilized to optimize image reconstruction and image quality. Since these algorithms will play an important role in the field of cardiac radiology, it is increasingly important for radiologists to be familiar with the potential applications of AI. The main focus of this article is to provide an overview of cardiac-related AI applications for CT and MRI studies, as well as non-imaging-based applications for reporting and image optimization.

Development and application of artificial intelligence in cardiac imaging

In this review, we describe the technical aspects of artificial intelligence (AI) in cardiac imaging, starting with radiomics, basic algorithms of deep learning and application tasks of algorithms, until recently the availability of the public database. Subsequently, we conducted a systematic literature search for recently published clinically relevant studies on AI in cardiac imaging. As a result, 24 and 14 studies using CT and MRI, respectively, were included and summarized. From these studies, it can be concluded that AI is widely applied in cardiac applications in the clinic, including coronary calcium scoring, coronary CT angiography, fractional flow reserve CT, plaque analysis, left ventricular myocardium analysis, diagnosis of myocardial infarction, prognosis of coronary artery disease, assessment of cardiac function, and diagnosis and prognosis of cardiomyopathy. These advancements show that AI has a promising prospect in cardiac imaging.

Radiomics in medical imaging-"how-to" guide and critical reflection

Radiomics is a quantitative approach to medical imaging, which aims at enhancing the existing data available to clinicians by means of advanced mathematical analysis. Through mathematical extraction of the spatial distribution of signal intensities and pixel interrelationships, radiomics quantifies textural information by using analysis methods from the field of artificial intelligence. Various studies from different fields in imaging have been published so far, highlighting the potential of radiomics to enhance clinical decision-making. However, the field faces several important challenges, which are mainly caused by the various technical factors influencing the extracted radiomic features.The aim of the present review is twofold: first, we present the typical workflow of a radiomics analysis and deliver a practical "how-to" guide for a typical radiomics analysis. Second, we discuss the current limitations of radiomics, suggest potential improvements, and summarize relevant literature on the subject.

Applying Artificial Intelligence to Mitigate Effects of Patient Motion or Other Complicating Factors on Image Quality

Artificial intelligence, particularly deep learning, offers several possibilities to improve the quality or speed of image acquisition in magnetic resonance imaging (MRI). In this article, we briefly review basic machine learning concepts and discuss commonly used neural network architectures for image-to-image translation. Recent examples in the literature describing application of machine learning techniques to clinical MR image acquisition or postprocessing are discussed. Machine learning can contribute to better image quality by improving spatial resolution, reducing image noise, and removing undesired motion or other artifacts. As patients occasionally are unable to tolerate lengthy acquisition times or gadolinium agents, machine learning can potentially assist MRI workflow and patient comfort by facilitating faster acquisitions or reducing exogenous contrast dosage. Although artificial intelligence approaches often have limitations, such as problems with generalizability or explainability, there is potential for these techniques to improve diagnostic utility, throughput, and patient experience in clinical MRI practice.

Radiomic Analysis of Native T1 Mapping Images Discriminates Between MYH7 and MYBPC3-Related Hypertrophic Cardiomyopathy

BACKGROUND

The phenotype via conventional cardiac MRI analysis of MYH7 (β-myosin heavy chain)- and MYBPC3 (β-myosin-binding protein C)-associated hypertrophic cardiomyopathy (HCM) groups is similar. Few studies exist on the genotypic-phenotypic association as assessed by machine learning in HCM patients.

PURPOSE

To explore the phenotypic differences based on radiomics analysis of T₁ mapping images between MYH7 and MYBPC3-associated HCM subgroups.

STUDY TYPE

Prospective observational study.

SUBJECTS

In all, 102 HCM patients with pathogenic, or likely pathogenic mutation, in MYH7 (n = 68) or MYBPC3 (n = 34) genes.

FIELD STRENGTH/SEQUENCE

Cardiac MRI was performed at 3.0T with balanced steady-state free precession (bSSFP), phase-sensitive inversion recovery (PSIR) late gadolinium enhancement (LGE), and modified Look-Locker inversion recovery (MOLLI) T₁ mapping sequences.

ASSESSMENT

All patients underwent next-generation sequencing and Sanger genetic sequencing. Left ventricular native T₁ and LGE were analyzed. One hundred and fifty-seven radiomic features were extracted and modeled using a support vector machine (SVM) combined with principal component analysis (PCA). Each subgroup was randomly split 4:1 (feature selection / test validation).

STATISTICAL TESTS

Mann-Whitney U-tests and Student's t-tests were performed to assess differences between subgroups. A receiver operating characteristic (ROC) curve was used to assess the model's ability to stratify patients based on radiomic features.

RESULTS

There were no significant differences between MYH7- and MYBPC3-associated HCM subgroups based on traditional native T₁ values (global, basal, and middle short-axis slice native T₁ ; P = 0.760, 0.914, and 0.178, respectively). However, the SVM model combined with PCA achieved an accuracy and area under the curve (AUC) of 92.0% and 0.968 (95% confidence interval [CI]: 0.968-0.971), respectively. For the test validation dataset, the accuracy and AUC were 85.5% and 0.886 (95% CI: 0.881-0.901), respectively.

DATA CONCLUSION

Radiomic analysis of native T₁ mapping images may be able to discriminate between MYH7- and MYBPC3-associated HCM patients, exceeding the performance of conventional native T₁ values.

LEVEL OF EVIDENCE

3 TECHNICAL EFFICACY STAGE: 2 J. MAGN. RESON. IMAGING 2020;52:1714-1721.

Characterization of interstitial diffuse fibrosis patterns using texture analysis of myocardial native T1 mapping

Background

The pattern of myocardial fibrosis differs significantly between different cardiomyopathies. Fibrosis in hypertrophic cardiomyopathy (HCM) is characteristically as patchy and regional but in dilated cardiomyopathy (DCM) as diffuse and global. We sought to investigate if texture analyses on myocardial native T₁ mapping can differentiate between fibrosis patterns in patients with HCM and DCM.

Methods

We prospectively acquired native myocardial T₁ mapping images for 321 subjects (55±15 years, 70% male): 65 control, 116 HCM, and 140 DCM patients. To quantify different fibrosis patterns, four sets of texture descriptors were used to extract 152 texture features from native T₁ maps. Seven features were sequentially selected to identify HCM- and DCM-specific patterns in 70% of data (training dataset). Pattern reproducibility and generalizability were tested on the rest of data (testing dataset) using support vector machines (SVM) and regression models.

Results

Pattern-derived texture features were capable to identify subjects in HCM, DCM, and controls cohorts with 202/237(85.2%) accuracy of all subjects in the training dataset using 10-fold cross-validation on SVM (AUC = 0.93, 0.93, and 0.93 for controls, HCM and DCM, respectively), while pattern-independent global native T₁ mapping was poorly capable to identify those subjects with 121/237(51.1%) accuracy (AUC = 0.78, 0.51, and 0.74) (P<0.001 for all). The pattern-derived features were reproducible with excellent intra- and inter-observer reliability and generalizable on the testing dataset with 75/84(89.3%) accuracy.

Conclusion

Texture analysis of myocardial native T₁ mapping can characterize fibrosis patterns in HCM and DCM patients and provides additional information beyond average native T₁ values.

Detection of Myocardial Tissue Alterations in Hypertrophic Cardiomyopathy Using Texture Analysis of T2-Weighted Short Inversion Time Inversion Recovery Magnetic Resonance Imaging

OBJECTIVE

The aim of this study was to evaluate the usefulness of texture analysis of T2-weighted short inversion time inversion recovery (T2-STIR) for detecting myocardial tissue alterations in hypertrophic cardiomyopathy (HCM).

METHODS

Twenty patients with HCM and 11 controls were examined. Texture analysis was performed for the hypertrophied regions with and without and abnormal hyperintensity (AHI) and for the interventricular septum of the controls on T2-STIR. T2 mapping was performed to measure myocardial T2 values.

RESULTS

A gray-level nonuniformity value of 64.7 was the best discriminator between patients and controls with an area under the curve of 0.93 on a receiver operating characteristic curve. T2 values did not differ between them. The gray-level nonuniformity was significantly smaller in AHI regions than in the hypertrophied regions without AHI in HCM patients.

CONCLUSIONS

Texture analysis is useful for quantitatively detecting myocardial tissue altenations, including AHI, associated with HCM on T2-STIR.

Clinical Importance of Myocardial T2 Mapping and Texture Analysis

Late gadolinium enhancement (LGE) magnetic resonance imaging (MRI) is valuable for diagnosis and assessment of the severity of various myocardial diseases owing to its potential to visualize myocardial scars. T₁ mapping is complementary to LGE because it can quantify the degree of myocardial fibrosis or edema. As such, T₁-weighted imaging techniques, including LGE using an inversion recovery sequence, contribute to cardiac MRI. T₂-weighted imaging is widely used to characterize the tissue of many organs. T₂-weighted imaging is used in cardiac MRI to identify myocardial edema related to chest pain, acute myocardial diseases, or severe myocardial injuries. However, it is difficult to determine the presence and extent of myocardial edema because of the low contrast between normal and diseased myocardium and image artifacts of T₂-weighted images and the lack of an established method to quantify the images. T₂ mapping quantifies myocardial T₂ values and help identify myocardial edema. The T₂ values are significantly related to the clinical symptoms or severity of nonischemic cardiomyopathy. Texture analysis is a postprocessing method to quantify tissue alterations that are reflected in the T₂-weighted images. Texture analysis provides a variety of parameters, such as skewness, entropy, and grey-scale non-uniformity, without the need for additional sequences. The abnormal signal intensity on T₂-weighted images or T₂ values may correspond to not only myocardial edema but also other tissue alterations. In this review, the techniques of cardiac T₂ mapping and texture analysis and their clinical relevance are described.

An integrated multi-objective whale optimized support vector machine and local texture feature model for severity prediction in subjects with cardiovascular disorder

PURPOSE

The left ventricle (LV) myocardium undergoes deterioration with the reduction in ejection fraction (EF). The analysis of its texture pattern plays a major role in diagnosis of heart muscle disease severity. Hence, a classification framework with co-occurrence of local ternary pattern feature (COALTP) and whale optimization algorithm has been attempted to improve the prediction accuracy of disease severity level.

METHODS

This analysis is carried out on 600 slices of 76 participants from Kaggle challenge that include subjects with normal and reduced EF. The myocardium of LV is segmented using optimized edge-based local Gaussian distribution energy (LGE)-based level set, and end-diastolic and end-systolic volumes were calculated. COALTP is extracted for two distance levels (d = 1 and 2). The t-test has been performed between the features of individual binary classes. The features are ranked using feature ranking methods. The experiments have been performed to analyze the performance of various percentages of features in each combination of bin for fivefold cross-validation. An integrated whale optimized feature selection and multi-classification framework is developed to classify the normal and pathological subjects using CMR images, and DeLong test has been performed to compare the ROCs.

RESULTS

The optimized edge embedded to level set has produced better segmented myocardium that correlates with R = 0.98 with gold standard volume. The t-test shows that texture features extracted from severe subjects with distance level "1" are more statistically significant with a p value (< 0.00004) compared to other pathologies. This approach has produced an overall multi-class accuracy of 75% [confidence interval (CI) 63.74-84.23%] and effective subclass specificity of 70% (CI 55.90-81.22%).

CONCLUSION

The obtained results show that the multi-objective whale optimized multi-class support vector machine framework can effectively discriminate the healthy and patients with reduced ejection fraction and potentially support the treatment process.

Image-Based Cardiac Diagnosis With Machine Learning: A Review

Cardiac imaging plays an important role in the diagnosis of cardiovascular disease (CVD). Until now, its role has been limited to visual and quantitative assessment of cardiac structure and function. However, with the advent of big data and machine learning, new opportunities are emerging to build artificial intelligence tools that will directly assist the clinician in the diagnosis of CVDs. This paper presents a thorough review of recent works in this field and provide the reader with a detailed presentation of the machine learning methods that can be further exploited to enable more automated, precise and early diagnosis of most CVDs.

Texture signatures of native myocardial T1 as novel imaging markers for identification of hypertrophic cardiomyopathy patients without scar

BACKGROUND

In patients with suspected or known hypertrophic cardiomyopathy (HCM), late gadolinium enhancement (LGE) provides diagnostic and prognostic value. However, contraindications and long-term retention of gadolinium have raised concern about repeated gadolinium administration in this population. Alternatively, native T₁ -mapping enables identification of focal fibrosis, the substrate of LGE. However HCM-specific heterogeneous fibrosis distribution leads to subtle T₁ -maps changes that are difficult to identify.

PURPOSE

To apply radiomic texture analysis on native T₁ -maps to identify patients with a low likelihood of LGE(+), thereby reducing the number of patients exposed to gadolinium administration.

STUDY TYPE

Retrospective interpretation of prospectively acquired data.

SUBJECTS

In all, 188 (54.7 ± 14.4 years, 71% men) with suspected or known HCM.

FIELD STRENGTH/SEQUENCE

A 1.5T scanner; slice-interleaved native T₁ -mapping (STONE) sequence and 3D LGE after administration of 0.1 mmol/kg of gadobenate dimeglumine.

ASSESSMENT

Left ventricular LGE images were location-matched with native T₁ -maps using anatomical landmarks. Using a split-sample validation approach, patients were randomly divided 3:1 (training/internal validation vs. test cohorts). To balance the data during training, 50% of LGE(-) slices were discarded.

STATISTICAL TESTS

Four sets of texture descriptors were applied to the training dataset for capture of spatially dependent and independent pixel statistics. Five texture features were sequentially selected with the best discriminatory capacity between LGE(+) and LGE(-) T₁ -maps and tested using a decision tree ensemble (DTE) classifier.

RESULTS

The selected texture features discriminated between LGE(+) and LGE(-) T₁ -maps with a c-statistic of 0.75 (95% confidence interval [CI]: 0.70-0.80) using 10-fold cross-validation during internal validation in the training dataset and 0.74 (95% CI: 0.65-0.83) in the independent test dataset. The DTE classifier provided adequate labeling of all (100%) LGE(+) patients and 37% of LGE(-) patients during testing.

DATA CONCLUSION

Radiomic analysis of native T₁ -images can identify ~1/3 of LGE(-) patients for whom gadolinium administration can be safely avoided.

LEVEL OF EVIDENCE

2 Technical Efficacy Stage: 2 J. Magn. Reson. Imaging 2020. J. Magn. Reson. Imaging 2020;52:906-919.

Machine learning based quantification of ejection and filling parameters by fully automated dynamic measurement of left ventricular volumes from cardiac magnetic resonance images

BACKGROUND

Although analysis of cardiac magnetic resonance (CMR) images provides accurate and reproducible measurements of left ventricular (LV) volumes, these measurements are usually not performed throughout the cardiac cycle because of lack of tools that would allow such analysis within a reasonable timeframe. A fully-automated machine-learning (ML) algorithm was recently developed to automatically generate LV volume-time curves. Our aim was to validate ejection and filling parameters calculated from these curves using conventional analysis as a reference.

METHODS

We studied 21 patients undergoing clinical CMR examinations. LV volume-time curves were obtained using the ML-based algorithm (Neosoft), and independently using slice-by-slice, frame-by-frame manual tracing of the endocardial boundaries. Ejection and filling parameters derived from these curves were compared between the two techniques. For each parameter, Bland-Altman bias and limits of agreement (LOA) were expressed in percent of the mean measured value.

RESULTS

Time-volume curves were generated using the automated ML analysis within 2.5 ± 0.5 min, considerably faster than the manual analysis (43 ± 14 min per patient, including ~10 slices with 25-32 frames per slice). Time-volume curves were similar between the two techniques in magnitude and shape. Size and function parameters extracted from these curves showed no significant inter-technique differences, reflected by high correlations, small biases (<10%) and mostly reasonably narrow LOA.

CONCLUSION

ML software for dynamic LV volume measurement allows fast and accurate, fully automated analysis of ejection and filling parameters, compared to manual tracing based analysis. The ability to quickly evaluate time-volume curves is important for a more comprehensive evaluation of the patient's cardiac function.

Myocardial Radiomics in Cardiac MRI

OBJECTIVE. The purpose of this article is to review the nascent field of radiomics in cardiac MRI. CONCLUSION. Cardiac MRI produces a large number of images in a fairly inefficient manner with sometimes limited clinical application. In the era of precision medicine, there is increasing need for imaging to account for a broader array of diseases in an efficient and objective manner. Radiomics, the extraction and analysis of quantitative imaging features from medical imaging, may offer potential solutions to this need.

Value of Computed Tomography Radiomic Features for Differentiation of Periprosthetic Mass in Patients With Suspected Prosthetic Valve Obstruction

Background:

We aimed to determine whether quantitative computed tomography radiomic features can aid in differentiating between the causes of prosthetic valve obstruction (PVO) in patients who had undergone prosthetic valve replacement.

Methods:

This retrospective study included 39 periprosthetic masses in 34 patients who underwent cardiac computed tomography scan from January 2014 to August 2017 and were clinically suspected as PVO. The cause of PVO was assessed by redo-surgery and follow-up imaging as standard reference, and classified as pannus, thrombus, or vegetation. Visual analysis was performed to assess the possible cause of PVO on axial and valve-dedicated views. Computed tomography radiomic analysis of periprosthetic masses was performed and radiomic features were extracted. The advantage of radiomic score compared with visual analysis for differentiation of pannus from other abnormalities was assessed.

Results:

Of 39 masses, there were 20 cases of pannus, 11 of thrombus, and 8 of vegetation on final diagnosis. The radiomic score was significantly higher in the pannus group compared with nonpannus group (mean, −0.156±0.422 versus −0.883±0.474; P <0.001). The area under the curve of radiomic score for diagnosis of pannus was 0.876 (95% CI, 0.731–0.960). Combination of radiomic score and visual analysis showed a better performance for the differentiation of pannus than visual analysis alone.

Conclusions:

Compared with visual analysis, computed tomography radiomic features may have added value for differentiating pannus from thrombus or vegetation in patients with suspected PVO.

Machine learning in cardiovascular magnetic resonance: basic concepts and applications

Machine learning (ML) is making a dramatic impact on cardiovascular magnetic resonance (CMR) in many ways. This review seeks to highlight the major areas in CMR where ML, and deep learning in particular, can assist clinicians and engineers in improving imaging efficiency, quality, image analysis and interpretation, as well as patient evaluation. We discuss recent developments in the field of ML relevant to CMR in the areas of image acquisition & reconstruction, image analysis, diagnostic evaluation and derivation of prognostic information. To date, the main impact of ML in CMR has been to significantly reduce the time required for image segmentation and analysis. Accurate and reproducible fully automated quantification of left and right ventricular mass and volume is now available in commercial products. Active research areas include reduction of image acquisition and reconstruction time, improving spatial and temporal resolution, and analysis of perfusion and myocardial mapping. Although large cohort studies are providing valuable data sets for ML training, care must be taken in extending applications to specific patient groups. Since ML algorithms can fail in unpredictable ways, it is important to mitigate this by open source publication of computational processes and datasets. Furthermore, controlled trials are needed to evaluate methods across multiple centers and patient groups.

Accuracy of Radiomics-Based Feature Analysis on Multiparametric Magnetic Resonance Images for Noninvasive Meningioma Grading

OBJECTIVE

Meningioma grading is relevant to therapy decisions in complete or partial resection, observation, and radiotherapy because higher grades are associated with tumor growth and recurrence. The differentiation of low and intermediate grades is particularly challenging. This study attempts to apply radiomics-based shape and texture analysis on routine multiparametric magnetic resonance imaging (MRI) from different scanners and institutions for grading.

METHODS

We used MRI data (T1-weighted/T2-weighted, T1-weighted-contrast-enhanced [T1CE], fluid-attenuated inversion recovery [FLAIR], diffusion-weighted imaging [DWI], apparent diffusion coefficient [ADC]) of grade I (n = 46) and grade II (n = 25) nontreated meningiomas with histologic workup. Two experienced radiologists performed manual tumor segmentations on FLAIR, T1CE, and ADC images in consensus. The MRI data were preprocessed through T1CE and T1-subtraction, coregistration, resampling, and normalization. A PyRadiomics package was used to generate 990 shape/texture features. Stepwise dimension reduction and robust radiomics feature selection were performed. Biopsy results were used as standard of reference.

RESULTS

Four statistically independent radiomics features were identified as showing the strongest predictive values for higher tumor grades: roundness-of-FLAIR-shape (area under curve [AUC], 0.80), cluster-shades-of-FLAIR/T1CE-gray-level (AUC, 0.80), DWI/ADC-gray-level-variability (AUC, 0.72), and FLAIR/T1CE-gray-level-energy (AUC, 0.76). In a multivariate logistic regression model, the combination of the features led to an AUC of 0.91 for the differentiation of grade I and grade II meningiomas.

CONCLUSIONS

Our results indicate that radiomics-based feature analysis applied on routine MRI is viable for meningioma grading, and a multivariate logistic regression model yielded strong classification performances. More advanced tumor stages are identifiable through certain shape parameters of the lesion, textural patterns in morphologic MRI sequences, and DWI/ADC variability.

Non-invasive prediction of genotype positive-phenotype negative in hypertrophic cardiomyopathy by 3D modern shape analysis

NEW FINDINGS

What is the central question of this study? Can impaired deformational indicators for genotype positive for hypertrophic cardiomyopathy in subjects that do not exhibit a left-ventricular wall hypertrophy condition (G+LVH-) be determined using non-invasive 3D echocardiography? What is the main finding and its importance? Using 3D-STE and modern shape analysis, peculiar deformational impairments can be detected in G+LVH- subjects that can be classified with good accuracy. Moreover, the patterns of impairment are located mainly on the apical region in agreement with other evidence coming from previous biomechanical investigations.

ABSTRACT

We propose a non-invasive procedure for predicting genotype positive for hypertrophic cardiomyopathy (HCM) in subjects that do not exhibit a left-ventricular wall hypertrophy condition (G+LVH-); the procedure is based on the enhanced analysis of medical imaging from 3D speckle tracking echocardiography (3D-STE). 3D-STE, due to its low quality images, has not been used so far to detect effectively the G+LVH- condition. Here, we post-processed echocardiographic images exploiting the tools of modern shape analysis, and we studied the motion of the left ventricle (LV) during an entire cycle. We enrolled 82 controls, 21 HCM patients and 11 G+LVH- subjects. We followed two steps: (i) we selected the most impaired regions of the LV by analysing its strains; and (ii) we used shape analysis on these regions to classify the subjects. The G+LVH- subjects showed different trajectories and deformational attributes. We found high classification performance in terms of area under the receiver operating characteristic curve (∼90), sensitivity (∼78) and specificity (∼79). Our results showed that (i) G+LVH- subjects present important deformational impairments relative to healthy controls and (ii) modern shape analysis can efficiently predict genotype by means of a non-invasive and inexpensive technique such as 3D-STE.

Machine learning-based radiomics strategy for prediction of cell proliferation in non-small cell lung cancer

PURPOSE

To explore the feasibility and performance of machine learning-based radiomics classifier to predict the cell proliferation(Ki-67)in non-small cell lung cancer (NSCLC).

METHODS

245 histopathological confirmed NSCLC patients who underwent CT scans were retrospectively included. The Ki-67 proliferation index (Ki-67 PI) were measured within 2 weeks after CT scans. A lesion volume of interest (VOI) was manually delineated and radiomics features were extracted by MaZda software from CT images. A random forest feature selection algorithm (RFFS) was used to reduce features. Six kinds of machine learning methods were used to establish radiomics classifiers, subjective imaging feature classifiers and combined classifiers, respectively. The performance of these classifiers was evaluated by the receiver operating characteristic curve (ROC) and compared with Delong test.

RESULTS

103 radiomics features were extracted and 20 optimal features were selected using RFFS. Among the radiomics classifiers established by six machine learning methods, random forest-based radiomics classifier achieved the best performance (AUC = 0.776) in predicting the Ki-67 expression level with sensitivity and specificity of 0.726 and 0.661, which was better than that of subjective imaging classifiers (AUC = 0.625, P < 0.05). However, the combined classifiers did not improve the predictive performance (AUC = 0.780, P > 0.05), with sensitivity and specificity of 0.752 and 0.633.

CONCLUSIONS

The machine learning-based CT radiomics classifier in NSCLC can facilitate the prediction of the expression level of Ki-67 and provide a novel non-invasive strategy for assessing the cell proliferation.

Presurgical differentiation between malignant haemangiopericytoma and angiomatous meningioma by a radiomics approach based on texture analysis

PURPOSE

To assess whether a machine-learning model based on texture analysis (TA) could yield a more accurate diagnosis in differentiating malignant haemangiopericytoma (HPC) from angiomatous meningioma (AM).

MATERIALS AND METHODS

Sixty-seven pathologically confirmed cases, including 24 malignant HPCs and 43 AMs between May 2013 and September 2017 were retrospectively reviewed. In each case, 498 radiomic features, including 12 clinical features and 486 texture features from MRI sequences (T2-FLAIR, DWI and enhanced T1WI), were extracted. Three neuroradiologists independently made diagnoses by vision. Four Support Vector Machine (SVM) classifiers were built, one based on clinical features and three based on texture features from three MRI sequences after feature selection. The diagnostic abilities of these classifiers and three neuroradiologists were evaluated by receiver operating characteristic (ROC) analysis.

RESULTS

Malignant HPCs were found to have larger sizes, slighter degrees of peritumoural oedema compared with AMs (P<0.05), and more serpentine-like vessels. The AUC of the enhanced T1WI-based classifier was 0.90, significantly higher than that of T2-FLAIR-based or DWI-based classifiers (0.77 and 0.73). The AUC of the SVM classifier based on clinical features was 0.66, slightly but not significantly lower than the performances of 3 neuroradiologists (AUC=0.69, 0.70 and 0.73).

CONCLUSION

Machine-learning models based on clinical features alone could not provide a better diagnostic performance than that of radiologists. The SVM classifier built by texture features extracted from enhanced T1WI is a promising tool to differentiate malignant HPC from AM before surgery.

Identifying Ventricular Arrhythmias and Their Predictors by Applying Machine Learning Methods to Electronic Health Records in Patients With Hypertrophic Cardiomyopathy (HCM-VAr-Risk Model)

Clinical risk stratification for sudden cardiac death (SCD) in hypertrophic cardiomyopathy (HC) employs rules derived from American College of Cardiology Foundation/American Heart Association (ACCF/AHA) guidelines or the HCM Risk-SCD model (C-index ∼0.69), which utilize a few clinical variables. We assessed whether data-driven machine learning methods that consider a wider range of variables can effectively identify HC patients with ventricular arrhythmias (VAr) that lead to SCD. We scanned the electronic health records of 711 HC patients for sustained ventricular tachycardia or ventricular fibrillation. Patients with ventricular tachycardia or ventricular fibrillation (n = 61) were tagged as VAr cases and the remaining (n = 650) as non-VAr. The 2-sample ttest and information gain criterion were used to identify the most informative clinical variables that distinguish VAr from non-VAr; patient records were reduced to include only these variables. Data imbalance stemming from low number of VAr cases was addressed by applying a combination of over- and undersampling strategies. We trained and tested multiple classifiers under this sampling approach, showing effective classification. We evaluated 93 clinical variables, of which 22 proved predictive of VAr. The ensemble of logistic regression and naïve Bayes classifiers, trained based on these 22 variables and corrected for data imbalance, was most effective in separating VAr from non-VAr cases (sensitivity = 0.73, specificity = 0.76, C-index = 0.83). Our method (HCM-VAr-Risk Model) identified 12 new predictors of VAr, in addition to 10 established SCD predictors. In conclusion, this is the first application of machine learning for identifying HC patients with VAr, using clinical attributes. Our model demonstrates good performance (C-index) compared with currently employed SCD prediction algorithms, while addressing imbalance inherent in clinical data.

Texture analysis of myocardial infarction in CT: Comparison with visual analysis and impact of iterative reconstruction

OBJECTIVES

To compare texture analysis (TA) with subjective visual diagnosis of myocardial infarction (MI) in cardiac computed tomography (CT) and to evaluate the impact of iterative reconstruction (IR).

METHODS

Ten patients (4 women, mean age 68 ± 11 years) with confirmed chronic MI and 20 controls (8 women, mean age 52 ± 11 years) with no cardiac abnormality underwent contrast-enhanced cardiac CT with the same protocol. Images were reconstructed with filtered back projection (FBP) and with advanced modeled IR at strength levels 3-5. Subjective diagnosis of MI was made by three independent, blinded readers with different experience levels. Classification of MI was performed using machine learning-based decision tree models for the entire data set and after splitting into training and test data to avoid overfitting.

RESULTS

Subjective visual analysis for diagnosis of MI showed excellent intrareader (kappa: 0.93) but poor interreader agreement (kappa: 0.3), with variable performance at different image reconstructions. TA showed high performance for all image reconstructions (correct classifications: 94%-97%, areas under the curve: 0.94-0.99). After splitting into training and test data, overall lower performances were observed, with best results for IR at level 5 (correct classifications: 73%, area under the curve: 0.65).

CONCLUSIONS

As compared with subjective, nonreliable visual analysis of inexperienced readers, TA enables objective and reproducible diagnosis of chronic MI in cardiac CT with higher accuracy. IR has a considerable impact on both subjective and objective image analysis.

Texture analysis of cardiovascular magnetic resonance cine images differentiates aetiologies of left ventricular hypertrophy

AIM

To investigate whether unenhanced cardiovascular magnetic resonance (CMR) balanced steady state free precession (bSSFP) cine images could be analysed using textural analysis (TA) software to differentiate different aetiologies of disease causing increased myocardial wall thickness (left ventricular hypertrophy [LVH]) and indicate the severity of myocardial tissue abnormality.

MATERIALS AND METHODS

A mid short axis unenhanced cine frame of 216 patients comprising 50 cases of hypertrophic cardiomyopathy (HCM; predominantly Left ventricular outflow tract obstruction [LVOTO] subtype), 52 cases of cardiac amyloid (CA; predominantly AL: light chain subtype), 68 cases of aortic stenosis (AS), 15 hypertensive patients with LVH (HTN+LVH), and 31 healthy volunteers (HV) underwent TA of the CMR cine images (CMRTA) using TexRAD (TexRAD Ltd, Cambridge, UK). Among the HV, 16/31 were scanned twice to form a test-retest reproducibility cohort. CMRTA comprised a filtration-histogram technique to extract and quantify features using six parameters.

RESULTS

Test-retest analysis in the HV showed a medium filter (3 mm) was the most reproducible (intra-class correlation of 0.9 for kurtosis and skewness and 0.8 for mean and SD). Disease cohorts were statistically different (p<0.001) to HV for all parameters. Pairwise comparisons of CMRTA parameters showed kurtosis and skewness was consistently significant in ranking the degree of difference from HV (greatest to least): CA, HCM, LVH+HTN, AS (p<0.001). Similarly, mean, standard deviation, entropy, and mean positive pixel (MPP) were consistent in ranking degree of difference from HV: HCM, CA, AS and HTN+LVH.

CONCLUSION

Radiomic features of bSSFP CMR data sets derived using TA show promise in discriminating between the aetiologies of LVH.

Radiomic Analysis of Myocardial Native T1 Imaging Discriminates Between Hypertensive Heart Disease and Hypertrophic Cardiomyopathy

OBJECTIVES

This study sought to examine the diagnostic ability of radiomic texture analysis (TA) on quantitative cardiovascular magnetic resonance images to differentiate between hypertensive heart disease (HHD) and hypertrophic cardiomyopathy (HCM).

BACKGROUND

HHD and HCM are associated with increased left ventricular wall thickness (LVWT). Contemporary guidelines define HCM as LVWT ≥15 mm that is unexplained by other disease, which complicates diagnosis in cases of co-occurrences. Conventional global native T₁ mapping involves calculation of mean T₁ values as a surrogate for fibrosis. However, there may be differences in its spatial localization, such as diffuse and more focal fibrosis in HHD and HCM, respectively.

METHODS

This study identified 232 subjects (53 with HHD, 108 with HCM, and 71 control subjects) for TA who consecutively underwent free-breathing multislice native T₁ mapping. Four sets of texture descriptors were applied to capture spatially dependent and independent pixel statistics. Six texture features were sequentially selected with the best discriminatory capacity between HHD and HCM and were tested using a support vector machine (SVM) classifier. Each disease group was randomly split 4:1 (feature selection/test validation), in which the reproducibility of the pattern was analyzed in the test validation dataset.

RESULTS

The selected texture features provided the maximum diagnostic accuracy of 86.2% (c-statistic: 0.820; 95% confidence interval [CI]: 0.769 to 0.903) using the SVM. For the test validation dataset, the accuracy of the pattern remained high at 80.0% (c-statistic: 0.89; 95% CI: 0.77 to 1.00). Global native T₁, with an accuracy of 64%, separated HHD and HCM patients modestly (c-statistic: 0.549; 95% CI: 0.452 to 0.640).

CONCLUSIONS

Radiomics analysis of native T₁ images discriminates between HHD and HCM patients and provides incremental value over global native T₁ mapping.

Cardiac MRI Texture Analysis of T1 and T2 Maps in Patients with Infarctlike Acute Myocarditis

Purpose To assess the diagnostic potential of texture analysis applied to T1 and T2 maps obtained with cardiac MRI for the diagnosis of acute infarctlike myocarditis. Materials and Methods This prospective study from August 2012 to May 2015 included 39 participants (overall mean age ± standard deviation, 34.7 years ± 12.2 [range, 18-63 years]; mean age of women, 46.1 years ± 10.8 [range, 24-63 years]; mean age of men, 29.8 years ± 9.2 [range, 18-56 years]) from the Magnetic Resonance Imaging in Myocarditis (MyoRacer) trial with clinical suspicion of acute myocarditis and infarctlike presentation. Participants underwent biventricular endomyocardial biopsy, cardiac catheterization, and cardiac MRI at 1.5 T, in which native T1 and T2 mapping as well as Lake Louise criteria (LLC) were assessed. Texture analysis was applied on T1 and T2 maps by using a freely available software package. Stepwise dimension reduction and texture feature selection was performed for selecting features enabling the diagnosis of myocarditis by using endomyocardial biopsy as the reference standard. Results Endomyocardial biopsy confirmed the diagnosis of acute myocarditis in 26 patients, whereas 13 participants had no signs of acute inflammation. Mean T1 and T2 values and LLC showed a low diagnostic performance, with area under the curve in receiver operating curve analyses as follows: 0.65 (95% confidence interval [CI]: 0.45, 0.85) for T1, 0.67 (95% CI: 0.49, 0.85) for T2, and 0.62 (95% CI: 0.42, 0.79) for LLC. Combining the texture features T2 run-length nonuniformity and gray-level nonuniformity resulted in higher diagnostic performance with an area under the curve of 0.88 (95% CI: 0.73, 1.00) (P < .001) and a sensitivity and specificity of 89% [95% CI: 81%, 93%] and 92% [95% CI: 77%, 93%], respectively. Conclusion Texture analysis of T2 maps shows high sensitivity and specificity for the diagnosis of acute infarctlike myocarditis. © RSNA, 2018 Online supplemental material is available for this article.