الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 153 |  |  |
| | | from | 153 | | |
Abstract
Myocardial fibrosis (MF) is a characteristic of cardiac remodelling arising from cardiac myocyte damage or replacement with fibrous tissue. MF is associated with myocardial stiffness and a progressive loss of myocardial contractility before frank heart failure develops (78,200). The main pathology of MF is due to increased myofibroblast activity and excessive extracellular matrix deposition between cardiac myocytes, which impairs its normal contractility (201). MF could be reversible that is why the identification and quantification of myocardial fibrosis play a significant role in the diagnosis, prevention and prognosis of cardiac diseases (200).
One of the key methods of diagnosis of myocardial fibrosis is by cardiovascular magnetic resonance (CMR) with gadolinium contrast administration(201). This helps us to recognise the spatial location, size of the scar and late gadolinium enhancement (LGE) patterns that help to differentiate between ischaemic and non-ischaemic myocardial injury. Ischaemic myocardial damage causes LGE which takes subendocardial or transmural distribution. However, non-ischaemic myocardial damage usually takes the sub-epicardial, mid-wall or at insertion points patterns (202–204) .
Cardiovascular magnetic resonance imaging (CMR) is a robust non-invasive imaging tool commonly used for assessing cardiac anatomy, function, and tissue characterization. A well-established tool for detecting regional myocardial fibrosis associated with adverse cardiovascular outcomes is CMR LGE (200) .
The measurement of myocardial function by left ventricular ejection fraction (LVEF) plays an important role in the assessment and management of patients in cardiology. For the evaluation of myocardial contractility, LVEF can be insensitive as LVEF is dependent on other variables, including preload and afterload and determine global function without assessing regional function, therefore LVEF does not always recognise subtle, but important, early changes in LV systolic function. Early changes in myocardial contractility not detected by reduced LVEF may also have significant clinical effects (205) .
Hence the evaluation of myocardial contractility, using strain as a measure of deformation, was proposed as a novel method applied through several techniques (28,205). Strain could evaluate myocardial performance and detect early systolic dysfunctions before a reduction in EF below the normal reference range is seen (206).
One of the methods for assessing strain is using CMR. To date, CMR tagging has been considered the gold standard for strain measurement. It has its limitations because it requires the acquisition of specific sequences and its tag lines are liable to fade in diastole. CMR-feature tracking (FT) is a relatively new post-processing tool and can be applied retrospectively and offline to the routinely acquired CMR cine images (206).
Routine CMR examination to assess cardiac diseases include both cine images and LGE images (207) . LGE sequences are dependent on gadolinium contrast administration. Gadolinium has a favourable safety profile but requires an intravenous line, increases the scan time and is contraindicated in advanced renal impairment. Recently, gadolinium contrast accumulation in the brain has been identified as a possible hazard for concern (208).
Summary
109
Consequently, In this study CMR FT was tested to evaluate the subtle myocardial dysfunction caused by MF which is detected in LGE images in a cohort with preserved EF >50% with no visual regional wall motion abnormalities (RWMA) to see if CMR FT could avoid the need for contrast administration or be a possible gatekeeper for contrast administration where it is felt to be needed. This could save the time, cost and potentially avoid the need for contrast administration during CMR.
Material and method:
Study population:
In a retrospective study with institutional approval by the Barts Hospital Ethics Committee, 729 cases were recruited from Bart Bioresource registry (BBR) who had undergone clinical CMR imaging in the period between January 2015 to June 2018. Patients with a diagnosis of myocardial infarction (ischaemic LGE), myocarditis (non-ischaemic LGE) and a control group with normal scans were included. The exclusion criteria included any CMR exam with diagnoses of cardiomyopathy such as hypertrophic cardiomyopathy, cardiac amyloidosis, Fabray’s disease, arrhythmogenic ventricular cardiomyopathy and sarcoidosis. A total of 182 cases were excluded due to having LVEF less than 50% or gross visual regional wall motion abnormalities, leaving 547. A further 145 cases were excluded due to poor quality LGE images and poor cine images leaving 402 cases included in the analysis.
CMR image acquisition:
CMR scans were performed on a 1.5T and 3T scanner (Siemens Healthineers, Erlangen, Germany. Cine images for functional assessment and TT analysis were acquired in 2-, 3-, and 4-chamber views, and in short-axis (SAX) views using a steady-state free precession (SSFP) sequence during breath-holding and with ECG triggering.
LGE images were acquired in the same planes as the cine images approximately 10 min after gadolinium administration, using respiratory motion–corrected, free-breathing single-shot SSFP, averaged phase-sensitive inversion recovery images (MOCO-LGE)
Cine CMR image analysis:
CMR image analysis was performed using commercially available software (CVI42 prototype 5.11; Circle Cardiovascular Imaging, Calgary, Canada). LV endocardial and epicardial borders were contoured in the cine images (SAX and LAX) via the new machine learning tool that performs fully automated contouring across all the cine stack and all phases and allows determination of RV insertion points in SAX view and longitudinal LV extent in LAX series.
All automatic contours were checked and manually corrected for the unacceptable contours to allow optimization of the endocardial and epicardial borders for the tissue tracking (TT) algorithm taking in consideration to include papillary muscles and endocardial trabeculae in the LV cavity. TT algorithm express both the 2D and 3D peak global strain and strain rate values in addition to providing regional values based on the AHA 16-segment mode.
Summary
110
LGE images analysis:
For LGE quantification, quantitative analysis was done at the global and regional 16 segment model. It was run in the same CVI42 software, endocardial and epicardial contouring for the SAX stack was done manually. Quantitative tool full width half max (FWHM) was used for all cases. To derive the AHA 16 segments quantification, LV extent in 4ch LGE image and RV insertion points in SAX stack were identified manually.
Results:
The study included a total of 108 patients with myocardial infarction (MI), 130 patients with myocarditis, and 164 as control cases, all with LVEF>50% and no gross visual wall motion abnormalities. Male sex was predominant in LGE groups compared to the control group. Risk factors were also predominant in MI group compared to the others.
MI group were older. LV and RV phenotypic parameters were generally within the normal reference range and no significant differences were noted between the groups except for LV mass in myocarditis which was higher than other groups.
Global CMR-TT strain parameters showed significant differences between LGE group and the control group (non-LGE group) except for the torsion indices which were similar. The global strains (radial, circumferential and longitudinal) correlated with the amount of LGE.
For regional CMR -FT strain parameters, 3808 segments from MI and myocarditis groups are subdivided into two categories: LGE segments (1338) and non-LGE segments (2470). Under each group, there is subgrouping by the number of AHA segment. Predominant segments with LGE show differences in strain parameters between LGE and non LGE segments.
LGE was prevalent in the basal and mid infro-lateral segments (AHA segments 4,5,10,11). The amount of LGE in the MYO group was higher than the MI group
Global strain parameters’ predictive ability:
Univariate analysis:
1- Prediction for presence or absence of LGE:
Our cohort involved 238 cases in LGE group vs 164 cases in non LGE group. the univariate logistic regression revealed that the best predictor for presence of LGE are the diastolic strain rate parameters which give ROC > 0.7.
2- Prediction for type of LGE:
The LGE group which include 108 cases with ischaemic LGE pattern and 130 non-ischaemic LGE pattern was included in the analysis which showed that SAX derived strain parameters (GRS, GCS) were able to discriminate between two types of LGE patterns. The lower values go with predicting the non-ischaemic LGE pattern.
Univariate CMR parameters + covariate analysis models:
The covariate models (common clinical and phenotypic LV parameters) also have acceptable discrimination for the presence and type of LGE and myocardial mass alone
Summary
111
has high predictive ability. With addition of best predictors of CMR strain parameters to the covariate models, they could reclassify the LGE cases from the control.
diaGRSR_SAX is the variable that improves discrimination for the presence of LGE the most, increasing the ROC area from 0.903 to 0.909 (p value=0.092) and NRI is 0.3545(p=0.00037).
for prediction of LGE type, There is no improvement in the ROC area when GRS_SAX or GCS_SAX were added to the models (p=0.803, p=0.858 respectively). NRI for GCS is 0.0523(p=0.68701)
Regional strain parameters prediction ability:
Univariate analysis:
1- Prediction for presence or absence of LGE:
Only the diseased group involved in the regional analysis from which 3808 segments were included in the analysis (1338 LGE segments and 2470 non LGE segments) for prediction of LGE presence or absence within each segment. The 21 CMR regional strain parameters show poor discriminatory ability. The predictive ability for all CMR variables measured by ROC curve were <0.7. After adjustment of the CMR strain parameters with the segment number, this improves their discriminatory power and two of SAX derived strain parameters have the best predictor ability (RS-SAX (ROC=0.725) and CS-SAX (ROC=0.725)) and one 3D strain parameter [CS-3D (ROC=0.724)].
Then the ability of best predictor strain variables to reclassify subjects when they added to the segment model was assessed using the continuous NRI. All of them show significant reclassification ability (fig 50).
2- Prediction of type of LGE:
Only LGE segments (775 non-ischaemic pattern and 563 ischaemic LGE pattern) were involved in this prediction model. The AUROC for all 21 CMR parameters were less than 0.7 and after adjustment of these models to the segment number, there is slight improvement in the discriminatory ability and only two of SAX derived strains had the highest predictive ability RS-SAX (AUROC=0.693), CS-SAX(AUROC=0.689). Both added accepted and statistically significant predictive power than that of the segment number. The lower strain values go to predict the non-ischaemic LGE pattern
Conclusion:
Peak global diastolic radial strain rate parameters is non-inferior to LGE contrast for prediction of subtle LV scar in cases of preserved LVEF and no regional wall motion abnormalities. In addition to, the SAX derived strains (radial and circumferential) could localize the segments with scar after adjustment with the segment number. But the strain parameters will not be able to differentiate the pattern and aetiology of underlying scar.
All these strain parameters despite they work modest in the statistical models, we can’t conclude they are surrogate for CMR contrast examination in our cohort study.