Published online Nov 30, 2020.
https://doi.org/10.3348/kjr.2019.0969
Radiomics of Non-Contrast-Enhanced T1 Mapping: Diagnostic and Predictive Performance for Myocardial Injury in Acute ST-Segment-Elevation Myocardial Infarction
Abstract
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.
INTRODUCTION
The severity of myocardial infarction (MI) is a determining factor of left ventricular (LV) dysfunction and long-term remodeling in post-ST-segment-elevation MI (STEMI) patients undergoing primary percutaneous coronary intervention (PCI) (1, 2). Cardiac magnetic resonance (CMR) imaging is regarded as the gold-standard noninvasive imaging technique for characterizing and quantifying myocardial tissue after acute and chronic MI injury. Despite the availability of several highly accurate predicting MR indicators (such as ejection fraction, infarct size, microvascular obstruction [MVO], transmurality, and myocardium rescue), the most appropriate parameter is still not definitively agreed on (3, 4, 5, 6).
The use of texture analysis (TA) is increasingly used to evaluate the intrinsic texture of tissues. By analyzing grayscale patterns and pixel interrelationships in the image, TA also detects tissue changes that are not easily perceptible to the naked eye (7). The use of TA in CMR imaging has recently been outlined in the context of MI, showing the potential for TA to detect small myocardial scars in cine images (8) and the potential to distinguish between acute MI and chronic MI (9).
CMR mapping techniques allow for the accurate characterization of the composition and viability of the myocardium after MI (10, 11) by quantitatively analyzing the ischemic injury, providing extra predictive indicators of remodeling and long-term mortality in the remote myocardium (12, 13) and infarcted core (14). Native T1 mapping provides a quantitative biomarker of myocardial intracellular and extracellular conditions without the use of intravenous contrast agents; further, it correctly demonstrates the myocardial area at risk following an acute ischemic event (15). In addition, new evidence suggests that native T1 mapping could enable the evaluation of the severity of injury and prediction of myocardial recovery (16) and can distinguish between reversible (i.e., edematous myocardium) and irreversible post-acute MI (i.e., necrotic myocardium as evaluated by late gadolinium enhancement [LGE]) without the use of contrast (17).
The incremental value of TA compared to T1 mapping in the assessment of myocardial injury is uncertain. Therefore, our study aimed to assess the diagnostic and predictive potential of T1 mapping acquired from CMR imaging based on TA for the evaluation of myocardial injury severity after PCI in patients with STEMI.
MATERIALS AND METHODS
Study Patients
The study was approved by the local ethics committee (No. 2019PS071J) and the written informed consent requirement was waived. We retrospectively included patients with STEMI who were first admitted from March 2018 to September 2019 and received PCI. The patients had CMR imaging at three days to seven days and at six months follow-up after PCI. The exclusion criteria included the presence of a previous MI, complications in the hospital (death, reinfarction, and clinical instability), poor CMR image quality, and incomplete CMR imaging data. Relevant data on the clinical history of all patients were collected prospectively, including demographics, hemodynamic, angiographic, and electrocardiographic information.
CMR Imaging Protocol
Imaging was performed using a Philips 3T MR scanner (Ingenia, Philips Healthcare). Two-chamber, four-chamber, and LV short-axis (SA) cine images were obtained using two-dimensional balanced steady-state free procession. The LV SA covered the entire left ventricle. The native T1 mapping scan was performed before the contrast injection, and the modified look-locker sequence was used to scan the base, the middle, and the apex. The acquisition mode was 3-(3)-3-(3)-5. An LGE scan was performed 15–20 minutes after the first contrast injection intravenously with 0.2 mmol/kg gadolinium contrast agent (Omniscan, GE Healthcare). The phase-sensitive inversion recovery (PSIR) sequence was used to cover the entire left ventricle on the SA. The CMR imaging parameters are listed in Table 1.
Table 1
CMR Scanning Parameters
CMR Imaging Analysis
Cardiac data were evaluated using CVI software (version 5.9.1, Circle Cardiovascular Imaging Inc.). Cardiac function was analyzed in cine images, including LV ejection fraction, LV end-diastolic volume, LV end-systolic volume, and segmental longitudinal strain (SLS). Segment analyses of SA images were conducted using the American Heart Association (AHA) proposed 16-segment model, and contours were drawn for the endocardium and the epicardium. SA images were split into six equiangular segments with the right ventricular-LV anterior intersection as a reference point. The measurements of the segments were evaluated as follows: peak SLS, native T1 value, and LGE percentage. The peak systolic longitudinal strain during the entire cardiac cycle was defined as the peak negative value (18). Segments with a change in LV SLS more than the 75th percentile were described as a favorable recovery of myocardial contractility. Segment T1 values obtained from SA native T1 maps were subject to strict and comprehensive quality control as previously described (Fig. 1) (16).
Fig. 1
Image shows the three different sections (base, mid-cavity, apex) of one post-ST-segment-elevation myocardial infarction patient.
From top to bottom, numbers within the panels indicate LGE, native T1 mapping, ROI segmentation, and feature extraction. The infarct segment showed high T1 values compared to the adjacent and remote segments, as demonstrated in the bull's eye maps of T1 values and LGE. A total of 1088 segments were divided into a training set and test set at a random ratio of 7:3 for further analysis. LGE = late gadolinium enhancement, ROI = region of interest
The infarcted myocardium size was measured on the PSIR image, and the infarcted myocardium was defined as a range larger than the mean signal of the normal myocardial region of interest (ROI) + 5 standard deviations (SDs). The segments in end-diastole were graded by peak LGE transmurality on a 5-point scale: 0 = 0%, 1 = 1–25%, 2 = 26–50%, 3 = 51–75%, and 4 = 76–100% (19). Segments that had LGE > 25% were defined as infarcted, while segments with no LGE and 180° from the infarction were defined as remote (20). Segments contiguous to the infarct and LGE < 25% were described as adjacent. Infarcted segments with persistent LGE ≥ 50% on 6-month scans were defined as irreversibly injured. MVO was defined by a low signal core in the MI area of the LGE image. The visual presence of low-signal core on T1 maps was detected for MVO (21).
Texture Feature Extraction and Dimension Reduction
TA was conducted using a free and open-source package (MaZda, version 4.6; Institute of Electronics, Technical University of Lodz) (22, 23) as described previously (24, 25). Three-section T1 maps (base, midventricular, and apical SA) automatically created by Philips 3T MR scanner were exported for further analysis as single Digital Imaging and Communications in Medicine images. ROIs were delineated twice in a subset of 20 subjects by the same reader to test intra-observer reproducibility of texture features and ROIs were delineated by a second radiologist to test inter-observer reproducibility. The left myocardium was divided into 16 segments based on AHA segments. To prevent partial volume effects, the trabeculated layer and epicardial boundary were carefully excluded, and visual artifacts were excluded from the area of interest (Fig. 1). Six subset texture characteristics were obtained separately (Supplementary Table 1), including a total of 279 texture characteristics.
The patients were split into training and test sets in the ratio of 7:3. Intra-class correlation coefficients (ICCs) were performed to assess intra- and inter-observer reproducibility and all features with an ICC < 0.75 were excluded. Student's t test or Mann Whitney nonparametric test as appropriate, stepwise logistic regression, and the least absolute shrinkage and selection operator (LASSO) regression algorithm were used to select the strongest features, which were used to construct the radiomics signature with multivariable logistic regression analysis in the training set. The strongest features were obtained at the tuning parameter (λ) with the maximum area under the curve (AUC), and the tuning parameter with the standard error (1 SD) of the maximum AUC was selected when no significant difference was found between the AUCs at the two different λ values (Supplementary Fig. 1). In both training and test sets, the predictive performance of radiomics features was evaluated by receiver operating characteristics analysis.
Statistical Analysis
All statistical analyses were implemented using R (version 3.5.1, R Foundation for Statistical Computing) with RStudio (version 1.0.136, RStudio). In the multivariate logistic regression model, the likelihood ratio test with backward step-down selection was applied. The LASSO regression analysis was used with a 10-fold cross-validation tuning of the penalty parameter on the basis of the maximum AUC. Comparisons of the predictive performances among the T1 values, optimal radiomics signature, and the combined radiomics model were performed using the Delong test. Categorical variables are presented as frequency (percentage), normally distributed continuous variables as the mean ± SD and other data as median (interquartile range [IQR]). The significance level was set at 0.05, and two-sided tests were used.
RESULTS
Study Population
The patient characteristics are given in Table 2. Overall, 77 patients were recruited, nine of whom were excluded for the following reasons: claustrophobia (n = 3), scan termination (n = 2), and poor CMR image quality of T1 maps or LGE images (n = 4). Of the remaining 68 patients with baseline CMR in one week, 40 patients had a follow-up scan at six months. Reasons for not returning for the follow-up CMR scan were withdrawal from the research (n = 25) and death (n = 3). Baseline and follow-up CMR measurements are summarized in Table 3.
Table 2
The Study Population's Basic Characteristics
Segmental Analysis
A total of 1088 segments in one week and 640 segments at six months were analyzed. Of all the baseline segments, 401 segments had LGE transmurality ≥ 75%, 45 segments had LGE transmurality ≥ 50% and < 75%, 36 segments had LGE transmurality ≥ 25% and < 50%, and 75 segments had LGE transmurality < 25%. MVO occurred in 32 (47%) patients, and the MVO mass was 3.0 ± 2.7 g. A total of 103 (9.5%) segments showed positive MVO, and 985 (90.5%) segments showed negative MVO. The mean baseline and follow-up SLS were −9.75% (IQR: −12.6% to −6.0%) and −10.3% (IQR: −13.0% to −6.6%), respectively, p < 0.001. The segments data showed no significant difference between the training sets and the test sets (Table 4).
Table 4
Comparisons of the Segment Characteristics in the Training Sets and the Test Sets
Value of T1-Mapping and TA in the Assessment of Baseline MVO
After the stepwise logistic regression model and cross-validating LASSO regression analysis (Supplementary Fig. 1), eight independent and the most important texture features derived from T1 mapping were selected for further statistical analyses, including one feature from the histogram (Perc. 90%), four features from the gray-level co-occurrence matrix [S(1,0) Entropy, S(0,1) Correlation, S(4,0) SumVarnc, S(5,0) DifEntrp], and three features from wavelets (WavEnLL_s-1, WavEnLL_s-2, WavEnLL_s-3).
In the training set, T1 values showed significant differences in the MVO-positive and MVO-negative groups, and all eight texture features showed significant differences between the two groups (Supplementary Table 2). The AUC for the presence of a hypointense core on T1 maps to detect MVO was 0.78 (95% CI: 0.72, 0.84) (Fig. 2). The combination of the eight features resulted in higher diagnostic performance, with an AUC of 0.86 (95% CI: 0.81, 0.91), compared to T1 values alone (Z = −2.729, p006), with an AUC of 0.72 (95% CI: 0.66, 0.79) in the training set (Fig. 2). Similar results were found in the test set; more details on the diagnostic performances are summarized in Table 5.
Fig. 2
Graph shows ROC analyses for positive MVO versus negative MVO in the training (A) and test (B) datasets.
ROC analysis indicates that combined T1 values and radiomics features had a higher accuracy for diagnosing segments with MVO compared to T1 values alone, visual assessment, and radiomics signature alone in the training or test datasets. AUC = area under the curve, MVO = microvascular obstruction, ROC = receiver operating characteristics
Table 5
Summary of the Diagnostic and Predictive Performances of CMR Parameters and Radiomics Signature in the Training and Test Datasets
Value of T1 Mapping and TA in the Assessment of Baseline LGE Peak Segmental Transmurality
Four independent and the most important texture features derived from T1 mapping were selected for further statistical analyses, including one feature from the histogram (mean) and three features from the gray-level co-occurrence matrix [S(0,1) Correlat, S(1,−1) SumEntrp, S(2,0) Correlat]. In the training set, T1 values showed significant differences in the LGE peak segmental transmurality groups; all four texture features showed significant differences between the two groups (Supplementary Table 3). The combination of T1 values and the four features resulted in higher diagnostic performance (Z = −4.603, p < 0.001), with an AUC of 0.84 (95% CI: 0.81, 0.87), compared to T1 values with an AUC of 0.79 (95% CI: 0.76, 0.83) (Fig. 3). For further analysis, when MVO positive segments were excluded from the two groups, the T1 value, radiomics signature and the combined parameters showed a basically equivalent diagnostic performance (Table 5). Similar results were showed in the test set; more details on the diagnostic performances are summarized in Table 5.
Fig. 3
Graph shows ROC analyses for peak LGE transmurality level = 0 versus peak LGE transmurality level ≥ 1 in the training (A) and test (B) datasets.
ROC analysis indicates that combined T1 values and radiomics features had the highest accuracy for diagnosing segments with LGE in the training or test datasets.
Predictive Value of T1 Mapping and TA in the Assessment of 6-Month Infarct Size
Three independent and the most important texture features derived from T1 mapping were selected for further statistical analyses, including one feature from the histogram (skewness) and two features from the gray-level co-occurrence matrix [S(0,1) Correlat, S(2,0) SumAverg]. In the training set, T1 values showed significant differences in the non-irreversible infarcted segments and irreversible infarcted segments, and all three texture features showed significant differences between the two groups (Supplementary Table 4). The baseline LGE percentage demonstrated the highest predictive performance for irreversible myocardial injury at six months with an AUC of 0.98 (95% CI: 0.97, 0.99) when compared to other parameters (Fig. 4). The combination of T1 values and the three features resulted in higher diagnostic performance, with an AUC of 0.84 (95% CI: 0.80, 0.88), a sensitivity of 0.86, and a specificity of 0.82, compared to T1 values (Z = −2.654, p007) or the radiomics signature (Z = −4.693, p < 0.001) (Fig. 4). The test set showed similar findings. More details on the predictive performances are summarized in Table 5.
Fig. 4
Graph shows ROC analyses for non-irreversible myocardial injury versus irreversible myocardial injury in the training (A) and test (B) datasets.
ROC analysis indicates that baseline LGE percentage predicted irreversible injury myocardial segments with the highest accuracy. Combined T1 values and radiomics features had a higher accuracy for predicting segments with irreversible myocardial injury compared to T1 values or radiomics signature alone in the training or test datasets.
Predictive Value of T1 Mapping and TA in the Assessment of the Recovery of SLS at 6-Month
Six independent and the most important texture features derived from T1 mapping were selected for further statistical analyses, including five features from the gray-level co-occurrence matrix [S(1,−1) Correlat, S(0,3) DifVarnc, S(0,4) SumVarnc, S(0,5) AngScMom, S(5,−5) Contrast] and one feature from the gray-level Run-length matrix (135dr_LngREmph). In the training set, T1 values showed no significant differences in the non-irreversible infarcted segments and irreversible infarcted segments, and all six texture features showed significant differences between the two groups (Supplementary Table 5). The combination of T1 values and the six features resulted in higher diagnostic performance, with an AUC of 0.76 (95% CI: 0.71, 0.81), a sensitivity of 0.68, and a specificity of 0.75, compared to T1 values (Z = −3.737, p < 0.001), with an AUC of 0.55 (95% CI: 0.49, 0.60) (Fig. 5). The test set demonstrated similar findings; more details on the predictive performances are summarized in Table 5.
Fig. 5
Graph shows ROC analyses for non-favorable SLS recovery versus favorable SLS recovery in the training (A) and test (B) datasets.
ROC analysis indicates that combined T1 values and radiomics features had the acceptable accuracy for predicting segments with favorable SLS recovery in the training or test datasets. SLS = segmental longitudinal strain
DISCUSSION
We investigated the added value of TA based on non-contrast-enhanced T1 mapping to diagnose myocardial injury and to predict myocardial functional recovery without the use of contrast agents. This study has several key findings: 1) radiomics features provide incremental diagnostic value for T1 values in distinguishing MVO-positive from MVO-negative myocardium, 2) combined radiomics and T1 values could detect the presence of MI at different extents of transmurality with higher diagnostic accuracy compared to T1 values alone, and 3) combined radiomics and T1 values predicted the irreversible segmental scar and recovery of LV SLS with good performance.
Several previous studies have investigated the diagnostic performance of native T1 mapping in myocardial injury. Research by Dall'Armellina et al. (16) on native T1 mapping in acute MI showed a strongly positive association between T1 values and LGE extent in segments without MVO. T1 mapping could be an important additional protocol for LGE and T2-weighted to recognize reversible myocardial injury (16). Our study demonstrated a good diagnostic performance (78–89%) of the visual assessment of T1 maps for MVO, consistent with the results of the previous study by Bulluck et al. (21) which indicated a fair diagnostic accuracy (79–81%) of MVO based on the “hypointense core” sign. In our study, although the single T1 values for MVO showed a fair diagnostic performance, the application of the radiomics signature demonstrated good performance, and the combination of T1 mapping and TA resulted in the highest accuracy for recognizing MVO.
The study by Dastidar et al. (26) demonstrated that segmental native T1 showed good diagnostic accuracy (AUC 0.83, CI 0.78–0.88) in acute MI for LGE transmurality (> 75%), including MVO. The sensitivity and specificity were 79% and 79%, respectively, which is generally in line with our results. TA also demonstrated higher diagnostic performance in identifying the presence of acute MI, which was not affected by MVO. Liu et al. (17) noted that acute native T1 mapping could differentiate reversible (edematous myocardium) and irreversible myocardial injury (infarction) at a certain cut-off value and that the volume of quantitatively irreversibly injured myocardium was in line with the LGE results. Our results showed that TA added value to T1 mapping to predict follow-up irreversible myocardial injury. T1 mapping was also recognized as a predictor of wall thickening (17) and functional recovery (16) in STEMI, which was similar to the result of this study. Radiomics signature demonstrated added value to T1 mapping to predict the recovery of LV SLS.
This study explored alternative approaches to TA to improve myocardial injury diagnosis without LGE. We did not directly associate our findings with histopathological analysis in this study, so we cannot define the specific association between the texture features and histopathological changes. The histopathological correlation of myocardial fibrosis and T1 mapping is well established (27, 28). Texture features are mathematical parameters derived from the pixel distribution that characterizes the structure underlying the image objects (29). MRI T1 images do not present microscopic details, but changes in histology may cause changes in texture that are suitable for analysis of texture. TA demonstrated good diagnostic performance for the detection of myocardial infarcted transmurality, regardless of the degree of peak transmurality. Our findings are generally in concordance with previous studies (8, 9).
The use of native T1 mapping for post-acute MI without use of contrast would be of major clinical use, in that it would not only be safer in patients with kidney impairment and would shorten the scan time (by avoiding contrast-based techniques) but would also potentially allow for an early accurate stratification of those acute patients in need of more aggressive treatment (30). Furthermore, TA based on T1 mapping, as it includes more information, may be used to assess the high risk of arrhythmia patients to implant ICD (31, 32).
There are several limitations to this study. First, additional research should be carried out to further clarify the impact of technical variables (including the field strength and sequence) on tissue inhomogeneity-related markers from TA analysis to T1 mapping. Second, ischemic vs. nonischemic MI cannot be distinguished by elevated segmental T1 values (26), and the application of TA to distinguish ischemic from nonischemic lesions is a potential research point. Third, TA should also be compared to other non-contrast approaches such as MR cine images (8) in future work, and the diagnostic accuracy of different combined techniques deserve further investigation. Fourth, the value of combined TA analysis and clinical factors such as age and myocardial enzymes in the long-term prognosis of STEMI patients requires further investigation.
In conclusion, combined radiomics parameters from non-contrast-enhanced T1 mapping and T1 values could obtain a higher diagnostic accuracy for acute myocardial injury, especially MVO. Radiomics of non-contrast-enhanced T1 mapping also provides incremental value in the prediction of irreversible myocardial damage and functional recovery at the 6-month follow-up. The radiomics approach of non-contrast-enhanced T1 mapping may have the potential to reduce the use of gadolinium contrast administration.
Supplementary Materials
The Data Supplement is available with this article at https://doi.org/10.3348/kjr.2019.0969.
List of Radiomics Features Used in This StudySupplementary Table 1
Texture feature selection using cross-validation LASSO regression analysis and the predictive performance for (A, B) microvascular obstruction presence, (C, D) peak late gadolinium enhancement transmurality ≥ 1, (E, F) irreversible myocardial injury, and (G, H) favorable segmental longitudinal strain recovery.Supplementary Fig. 1
Results of T1 Values and Texture Analysis in the Segmental MVO-Negative Group and MVO-Positive GroupSupplementary Table 2
Results of T1 Values and Texture Analysis in the Segmental Infarction Absence and Infarction Presence GroupsSupplementary Table 3
Results of T1 Values and Texture Analysis in the Irreversible Segments and Non-Irreversible SegmentsSupplementary Table 4
Results of T1 Values and Texture Analysis in the Segments of Non-Favorable and Favorable SLS RecoverySupplementary Table 5
The study was funded by the 345 Talent Project in Shengjing Hospital of China Medical University.
Conflicts of Interest:The authors have no potential conflicts of interest to disclose.
Acknowledgments
We thank Yan Guo and Jianqing Sun for their expert opinion and helpful comments.
References
-
Yamashita Y, Shiomi H, Morimoto T, Yaku H, Furukawa Y, Nakagawa Y, et al. Cardiac and noncardiac causes of long-term mortality in ST-segment-elevation acute myocardial infarction patients who underwent primary percutaneous coronary intervention. Circ Cardiovasc Qual Outcomes 2017;10:e002790
-
-
Larroza A, Bodí V, Moratal D. Texture analysis in magnetic resonance imaging: review and considerations for future applications. In: Constantinides C, editor. Assessment of cellular and organ function and dysfunction using direct and derived MRI methodologies. London: IntechOpen; 2016.
-
-
Garg P, Broadbent DA, Swoboda PP, Foley JRJ, Fent GJ, Musa TA, et al. Acute infarct extracellular volume mapping to quantify myocardial area at risk and chronic infarct size on cardiovascular magnetic resonance imaging. Circ Cardiovasc Imaging 2017;10:e006182
-
-
Carrick D, Haig C, Rauhalammi S, Ahmed N, Mordi I, McEntegart M, et al. Prognostic significance of infarct core pathology revealed by quantitative non-contrast in comparison with contrast cardiac magnetic resonance imaging in reperfused ST-elevation myocardial infarction survivors. Eur Heart J 2016;37:1044–1059.
-
-
Dall'Armellina E, Piechnik SK, Ferreira VM, Si QL, Robson MD, Francis JM, et al. Cardiovascular magnetic resonance by noncontrast T1-mapping allows assessment of severity of injury in acute myocardial infarction. J Cardiovasc Magn Reson 2012;14:15
-
-
Liu D, Borlotti A, Viliani D, Jerosch-Herold M, Alkhalil M, De Maria GL, et al. CMR native T1 mapping allows differentiation of reversible versus irreversible myocardial damage in ST-segment-elevation myocardial infarction: an OxAMI Study (Oxford Acute Myocardial Infarction). Circ Cardiovasc Imaging 2017;10:e005986
-
-
Schulz-Menger J, Bluemke DA, Bremerich J, Flamm SD, Fogel MA, Friedrich MG, et al. Standardized image interpretation and postprocessing in cardiovascular magnetic resonance: Society for Cardiovascular Magnetic Resonance (SCMR) board of trustees task force on standardized post processing. J Cardiovasc Magn Reson 2013;15:35
-
-
Strzelecki M, Szczypinski P, Materka A, Klepaczko A. A software tool for automatic classification and segmentation of 2D/3D medical images. Nucl Instrum Methods Phys Res A 2013;702:137–140.
-
-
Baeßler B, Mannil M, Maintz D, Alkadhi H, Manka R. Texture analysis and machine learning of non-contrast T1-weighted MR images in patients with hypertrophic cardiomyopathy-preliminary results. Eur J Radiol 2018;102:61–67.
-
-
Kotu LP, Engan K, Eftestøl T, Ørn S, Woie L. Segmentation of scarred and non-scarred myocardium in LG enhanced CMR images using intensity-based textural analysis. Annu Int Conf IEEE Eng Med Biol Soc 2011;2011:5698–5701.
-
-
Eftestøl T, Woie L, Engan K, Kvaløy JT, Nilsen DWT, Ørn S. Texture analysis to assess risk of serious arrhythmias after myocardial infarction. Computing in Cardiology IEEE 2012;2012:365–368.
-
Publication Types
MeSH Terms
Figures
1 / 5
Tables
1 / 5
ORCID IDs
Funding Information
-
Shengjing Hospital
501100015226