New Insights in Strain Mechanics (LA, RA, and RV)

The purpose of this paper is to review the current status and literature surrounding left atrial, right atrial, and right ventricular strain. Advancements in chamber-specific strain software and taskforce consensus statements have helped overcome the previous limitations of reproducibility and inter-vendor variability. Strain has increasing utility due to its superior diagnostic sensitivity and independent prognostic value over traditional imaging assessments in a range of clinical conditions. The use of strain as a reliable and reproducible marker of cardiac function is most widely accepted in the assessment of left ventricular global longitudinal strain (GLS). However, strain can also be assessed in other cardiac chambers (left atrium (LA), right ventricle (RV), and right atrium (RA)). Consolidation and refinement of strain assessments in these other chambers have been achieved by chamber-specific software and uniform approaches to measurements. Strain accounts for the complex anatomy and physiology of these chambers and therefore holds sensitive diagnostic capacity. Current clinical applications are mainly in prognostication however utility is expanding specifically in LA strain, to identify and manage occult atrial fibrillation and in estimation of LV filling pressures. Further research is required to determine a universal approach in RV strain measurements and to improve technical capabilities in RA strain assessment.


Introduction
Strain denotes the degree of deformation of matter that occurs in response to an applied stress. Its measurement with ultrasound-initially with tissue Doppler imaging (TDI) and currently almost exclusively with two-dimensional (2D) speckle tracking echocardiography (STE)-has mainly been applied to the assessment of left ventricular (LV) function. The most widely adopted measurement is LV global longitudinal strain (GLS), which has become accepted as a robust and reproducible measurement of LV function [1]. The use of GLS for both detection of subclinical dysfunction and in the prediction of cardiovascular morbidity and mortality has led to its acceptance in guidelines (Table 1) [2, 3•, 4-7].
In addition, strain can also be measured in other cardiac chambers. While LV strain can be measured in the longitudinal, circumferential or radial dimensions, atrial and right ventricular (RV) strain have almost exclusively been measured by longitudinal shortening-which is quantified by a negative value. Although not having attained the same level of support in guidelines, increasing evidence supports the adoption of RV strain and atrial strain as clinical tools. This paper reviews the current status of left (LA) and right atrial (RA) and RV strain, encompassing assessment, normal values, technical considerations, and clinical applications.

Rationale for the assessment of LA function
Traditionally, imaging assessment of the LA has been restricted to LA size, with volumetric assessment having superseded the evaluation of anteroposterior dimensions, which enlarge non-uniformly [8]. However, in general, cardiac dysfunction precedes structural change, so a functional 1 3 marker would be helpful. While LA function can be assessed using Doppler (transmitral A wave and pulmonary venous atrial reversal), this is cumbersome. Likewise, the use of 2D imaging for the measurement of LA ejection fraction is constrained by geometrical assumptions, and while these can be avoided with 3D imaging, images in the far field may not be well-resolved. Atrial longitudinal strain measures the change in length of the atrial myocardium through the cardiac cycle [9•]. The use of LA strain as a functional parameter provides a useful prognostic tool for a range of cardiovascular pathologies [10]. Measurements of LA function can be used to assess LV filling pressure, assess the consequences of valve disease, and permit the detection of atrial cardiomyopathies, which help to define the risk of AF [11,12].

Atrial Physiology
LA function is divided into 3 phases: reservoir, conduit, and contractile [13]. The reservoir phase reflects the filling of the left atrium via pulmonary veins with a closed mitral valve. The conduit phase occurs during early diastole when the mitral valve opens and the LA empties into the LV-corresponding to the E wave (early transmitral flow). The contractile phase relates to the A wave (late transmitral flow) and represents left atrial contraction ( Fig. 1) [9•, 14]. LA function is affected by loading conditions and heart rate in all three phases.
LA reservoir strain (LARS) has prognostic value in a variety of cardiovascular diseases [10] and predicts morbidity and mortality in the general population [15]. The anatomy and thin walls of the LA make this chamber susceptible to atrial remodelling from pressure/volume overload or arrhythmic damage. LA filling during the reservoir phase results in stretching of the atrial wall, therefore pathologies that cause a reduction in atrial compliance have a direct effect on LA reservoir phasic function [16]. In summary, the degree of LA interstitial fibrosis correlates to LA reservoir function and this is most accurately evaluated by LA strain rather than alternative echocardiographic measurements [16].
LA contractile strain (LACS) and LA conduit strain (LAScd) have less clinical utility when compared to LARS and require more focused research to further explore their applications. LACS correlates to the active contraction of the atria and is affected by elevated LV filling pressures as well as reduced LA contractile function. LACS has predictive value in AF-reduced LACS is a predictor for the development of new AF [17], and preserved LACS predicts maintenance of sinus rhythm post DCR [18]. LACS may also help predict recurrence of AF post catheter ablation, even in patients with normal-sized LA [19].
LA conduit strain (LAScd) is an indirect marker of LV relaxation as it results from the pressure gradient between LA and LV with an open mitral valve. Impaired LAScd is associated with heart failure with preserved ejection fraction (HFpEF) [20] and diastolic dysfunction [21]. LACS has demonstrated prognostic value in specific populations such as patients with congenital aortic stenosis [22] and patients with ESRF [23], however further research is still required for wider application.

Technical Aspects of LA Strain Measurement
The EACVI/ASE/Industry Taskforce has released a consensus statement to standardise LA strain assessment and reduce inter-vendor variability [9•]. LA strain should be assessed using an optimised apical four-chamber view at a high frame rate (50-70 frames/min). Evidence supports the use of global rather than regional LA strain, and sampling over more of the LA wall probably provides less opportunity for sampling error than a single view. We use apical 4-and 2-chamber views, but not the apical long axis because of potential problems with inclusion of the pulmonary veins and left atrial appendage [9•]. Manual correction to endocardial tracking is often needed due to the thin LA and mobile interatrial septum, which may provide problems for automated tracking [14]. The recent development of specific LA-strain software (e.g. AutoStrain LA and LA Automated Function Imaging) has further consolidated a uniform approach to LA-strain measurement and allows for greater accessibility of these measurements by non-expert operators [24].
Two temporal reference points may be used for tracking the LA border; QRS guided or P wave guided (Fig. 1). The net displacement with each (ie reservoir strain) is similar. Both options have specific considerations-for example QRS complex gating is more challenging in a bundle branch block and P-P gating is not possible in AF [9 •]. Current LA-strain software automatically selects the upslope of the R-wave as the surrogate for end-diastole and generates a template for endocardial tracing. The MASCOT-HIT study took 26 expert centres and showed that both reference points for LA-strain assessment were reproducible, but QRS-guided LA-strain was more feasible and had a shorter analysis time [24]. Normal values for LARS have been assessed in two meta-analyses [3 •, 25] ( Table 2). Variation of the "normal value" for LARS was influenced by differing ECG gating, sample size and vendor, as well as age, gender, and race. Increasing age correlated with deteriorating LA reservoir and conduit function [4] and this effect appeared to be more pronounced in women [26]. The normal value for LARS is ≈ − 38%; however, studies have shown that a LARS cutoff of ≈ > − 22% identifies risk of adverse outcomes [27,28]. This implies that LARS has a significant reserve.

Clinical Considerations
Clinical applications of LA strain are broad but are most relevant to atrial fibrillation (AF), HFpEF, and valvular heart disease; AF is accompanied by left atrial structural and electrical remodelling [29]. Assessing LA strain to evaluate LA function can aid in the prediction of AF in multiple clinical settings (Table 3) [30-32, 33•, 34-42, 43•]. In the Copenhagen City Heart Study [30], 4466 healthy participants underwent LA strain assessment and were followed for a median of 5.3 years for incident AF (which occurred in 4.3% of participants). Results showed a LA strain of < 23% was associated with a 6.8 increased risk of AF compared to those with LA strain ≥ 23%. Abnormal LARS and LACS were independent predictors of AF. Abnormal LARS (≤ 19%) is also a useful predictor of AF occurrence in patients with hypertension [44]. In patients after cryptogenic stroke, abnormal LA strain is predictive of future AF occurrence, demonstrating a potential clinical benefit for re-classifying some of these patients as having an atriopathy [33•, 34, 35]. A meta-analysis of 12 studies, involving 1025 patients post-radiofrequency ablation revealed that LARS was significantly lower pre-ablation in patients with recurrent AF compared with those who maintained sinus rhythm (15.7 ± 5.7% vs. 23.0 ± 7.0%, p = 0.016). LARS was independently associated with AF recurrence [41], potentially because it reflects atriopathy and fibrosis as the underlying substrates for AF. LA strain not only predicts AF recurrence after catheter ablation [45] but also remains an independent predictor of AF in patients even with a non-dilated LA [36,40]. This ability for strain to risk-stratify the recurrence of AF may aid in determining which patients will benefit from longer-term monitoring and/or anticoagulation [35].
HFpEF is an increasing phenotype of heart failure, now accounting for more than 50% of heart failure presentations [46]. LA strain can provide incremental information that facilitates the diagnosis of HFpEF. In 517 patients with HFpEF and elevated filling pressures, impairment in LARS (< 23%) was almost twice as common as increased LAVi (62.4% vs. 33.6%, p < 0.01) [47]. In sub-analyses of the large HFpEF management trials such as TOPCAT (Treatment Of Preserved Cardiac Function Heart Failure With an Aldosterone Antagonist Trial) and PARAMOUNT (Prospective comparison of ARNI with ARB on Management Of heart failUre with preserved ejectioN fracTion), LARS was consistently shown to be impaired in patients with HFpEF when compared with controls [48,49]. In PARAMOUNT, LARS was reduced in patients with HFpEF (48.4 ± 1.2% vs. controls 58.4 ± 2.1%, p < 0.001) and in TOPCAT a LARS of < 26% was detected in 52% of HFpEF patients [48,49].
Nonetheless, the recognition of HFpEF is based on multiple clinical and echocardiographic variables, and while the adoption of LA strain may facilitate this process, no criteria provide this diagnosis in isolation. Traditional echocardiographic assessment of HFpEF integrates pulmonary artery pressure, transmitral flow, tissue velocity, and maximum indexed left atrial volume (LAVi) [46,50]. In an attempt to simplify the diagnostic process, Reddy et al. proposed the "H2PEFF" score which accounts for echocardiographic and clinical parameters (E/e′ ratio > 9, pulmonary artery systolic pressure > 35 mm Hg, obesity, atrial fibrillation, age > 60 years, ≥ 2 antihypertensive agents) [51•]. These variables derive a composite score from 0 to 9 where the odds of HFpEF doubled for each 1-unit score increase. This score was  found to correctly discriminate unexplained dyspnea due to HFpEF from noncardiac dyspnea (with invasive haemodynamic exercise testing as the gold standard for the diagnosis of HFpEF). Another diagnostic algorithm is the ESC Heart Failure Association algorithm "HFA-PEFF" [52]: Pretest Assessment (P), Diagnostic workup with echocardiogram and natriuretic peptide (E), Functional testing in case of uncertainty (F), and Final etiological workup (F). In this diagnostic approach, a score of ≥ 5 is diagnostic of HFpEF whilst a score of ≤ 1 is unlikely to be HFpEF, scores 2-4 are indeterminate and require further testing. Both these scores have evolved the way the diagnosis of HFpEF is made from a binary approach to an estimation of likelihood of HFpEF. Despite this, there is still variability when the algorithms are applied to the same population with 41% of suspected HFpEF patients being classified differently by either score [53]. Of note, neither of these scores utilise the capabilities of LA strain. The importance of assessing LA filling pressure is not restricted to HFpEF. The current diagnostic algorithm for diastolic dysfunction is heavily dependent on measurement of LAVi and E/e′, but this assessment often provides equivocal findings [54]. Part of the problem is that although elevated filling pressures lead to left atrial dilatation and eventual remodelling with stretching of atrial cardiomyocytes [55], this process is nonspecific (often caused by AF) and the process of reverse remodelling is incomplete and may be delayed. Assessment of LA function (rather than just structure) with left atrial strain has been shown to be a sensitive tool in its ability to discriminate between all stages of diastolic dysfunction [54]. LARS improves the feasibility and diagnostic accuracy of the 2016 ASE/EACVI diastolic algorithm in patients with HFpEF [56]. LA strain measurements allow for accurate classification of diastolic dysfunction with a LARS of > 35% correlating to patients with normal diastolic function with an accuracy of 72% [57]. Furthermore, abnormal LARS of ≤ 19% identified patients with grade 3 diastolic dysfunction to an accuracy of 95% [57].
Valvular heart disease is associated with impaired LA strain, which also has prognostic significance. Mitral regurgitation (MR) of any aetiology is associated with reduced LA (reservoir and contractile) strain [58]. Sugimoto et al. found that secondary MR when compared to primary MR at similar severity, has a more significantly impaired LA strain profile [58]. In the same study, LARS > 16% at peak exercise was associated with 3-year event-free survival [58]. In asymptomatic patients with moderate MR, a reduced LARS was associated with the development of subsequent cardiovascular events [59]. Pre-operative mitral valve surgery patients with severe primary MR demonstrated reductions in LARS (< 21%) as an independent predictor of mortality, HF, and functional capacity [60]. Future applications of LARS could be to optimise timing of mitral valve surgery to improve post-operative outcomes.

Rationale for the Assessment of RV Function
The assessment of RV function is a cornerstone in the assessment of pulmonary hypertension, heart failure, and valvular abnormalities. Unfortunately, due to its retrosternal position, crescent-shaped geometry, and load dependency, echocardiographic assessment of the RV is challenging [61]. RV function is also complex and occurs via three mechanisms: longitudinal shortening with movement of tricuspid annulus towards the apex, radial motion with movement of free wall inward producing a bellows effect, and anteroposterior shortening from free wall movement over the septum [62]. The current gold standard for assessment of RV function is RV volumetric analysis and ejection fraction (RVEF) using CMR. However, accessibility to CMR remains restricted in some places, primarily due to cost, accessibility, and contra-indications in patients with non-compatible devices.
Traditional quantitative echocardiographic markers such as tricuspid annular plane systolic excursion (TAPSE) and RV fractional area change (FAC) are subject to translational motion and have limited prognostic and diagnostic yield [63]. Additionally, TAPSE is vulnerable to overestimation with translational motion and is only partially representative as it evaluates just one segment of the RV. RV FAC is limited by image quality and relies heavily on the correct identification of the endocardial border [64]. In this context, novel methods such as RV strain and three-dimensional echocardiography (3DE) are emerging as promising additions to traditional parameters.
The angle independence of RV strain enables it to measure longitudinal function in the entire RV and RV free wall. As in the atria and LV, RV strain permits the detection of subclinical RV dysfunction while conventional markers of RV function are still normal [63]. RV longitudinal strain is a reproducible and highly sensitive marker of RV dysfunction [65], which is of value in cardiomyopathies, cardiac amyloidosis, and cancer [66]. 3D echocardiography (3DE) also accounts for the complex structure of the RV and overcomes limitations of foreshortened images and geometric assumptions [67]. Accuracy and reproducibility have been well established in comparison to CMR [68] and prognostic benefit over standard 2D echocardiography has also been demonstrated [69]. While RV strain and 3D-RVEF are load dependent (like any RVejection phase parameter), 3DE may be more susceptible to confounding by altered loading conditions [70]. Nonetheless, both appear to have independent prognostic value in conditions such as PH, heart failure, and congenital heart disease [71][72][73]. 3DE and RV strain correlate well with CMR RVEF and are good predictors of RV dysfunction, superior to current standard 2D parameters [74].

RV Physiology
The RV is exquisitely sensitive to increased pulmonary arterial (PA) loading and the ability for the RV to adapt to these changes in conditions has significant prognostic implications [75]. RV-PA coupling is a representation of how RV contractility matches RV afterload. An increase in RV afterload should result in an associated increase in RV contractility (in which case the RV and PA are "coupled"), and adverse outcomes are associated with RV-PA "uncoupling". The gold standard assessment of RV-PA coupling is from invasive multi-beat RV pressure-volume evaluation of arterial and ventricular elastance. A non-invasive surrogate of RV-PA coupling is the ratio of TAPSE to PASP, with a TAPSE/ PASP cutoff of 0.31 mm/mmHg identifying RV/PA uncoupling and poor prognosis [76]. 3DE can be used to determine a RVEF/PASP ratio, which has been used as a novel marker of RV-PA coupling [77,78]. Indexing strain assessment of RV function against PASP (via RV-GLS/PASP and RV-FWS/PASP) also provides a representation of RV-PA coupling and consistently demonstrates an increased risk of mortality [79].

Technical Aspects of RV Strain Measurement
RV strain should be assessed in a dedicated RV-focused apical four-chamber view. The RV should be completely visualised with optimization of the RV-free wall [63]. Standard apical four-chamber views will give a smaller strain value than a focused-RV view [80•]. The RV strain can be assessed either as RV-GLS (a 6-segment model involving the intraventricular septum (IVS) and free-wall), or just as RV free-wall strain (RV-FWS: 3-segment model) [80•]. RV free-wall strain is larger in magnitude than RV GLS. The measurement of either RV-GLS or RV-FWS may be appropriate, depending on the circumstance. For example, questions purely relating to the right-sided circulation (e.g. relating to RV responses to pulmonary hypertension) are best addressed by RV-FWS, while questions pertaining to RV function (e.g. response to tricuspid regurgitation or RV infarction) might include assessment of the septum. Nonetheless, measurement of RV-FWS has been supported by the EACVI/ASE/Industry Task Force and has a wider consensus on established normal values [9•, 80•, 81].
A recent meta-analysis of normal values for RV strain reviewed 788 articles to include 4439 healthy subjects. The analysis defined normal reference values with a pooled mean and lower limit of normal with 95% confidence interval for RVFWLS of − 26.9% (− 28.0%, − 25.9%) and RVGLS of − 23.4% (− 24.2, − 22.6%) respectively [82].
Recently introduced dedicated RV strain software may overcome some of the technical challenges of using the LV package to assess the RV. A study using CMR as the gold standard found that traditional parameters for RV function (TAPSE, FAC, RV S') and contemporary RV strain all provided good correlation with RV ejection fraction. However, RV-FWS provided the strongest diagnostic accuracy to predict an impaired RVEF (< 45%) with a high specificity and sensitivity [83].
One of the limitations of 2D strain in general is transition of speckles to outside the imaging plane during systole. This is problematic with assessment of the RV-free wall, and a reason to consider 3D strain. Unfortunately, despite advances in parallel processing, this modality necessitates trade-offs in temporal and spatial resolution that may lead to under-sampling and compromise sensitivity, as well as introducing artefact. For this reason, we do not employ this modality in the clinical routine.

Clinical Applications of RV Strain
RV strain assessment provides value in pulmonary arterial hypertension (PAH), cardiomyopathies, and valvular heart disease.
In PAH, the status of the RV is a major determinant of outcome. RV strain may be impaired whilst conventional parameters for RV function are still normal-a situation described as subclinical RV dysfunction [84]. Abnormal RV-FWS in PAH is a powerful predictor of future cardiovascular events over a 4-year follow-up [85]. Furthermore, unlike TAPSE and fractional area change, RV FWS was an independent predictor for all-cause mortality in patients with PAH [86]. The magnitude of change in RV strain is also clinically relevant-a ≥ 5% reduction in RV FWS has a > sevenfold lower mortality risk over a 4-year period [87].
In left heart failure (HF), RV strain also plays an important role in prognostic stratification. RV strain may deteriorate early in LHF if LA pressure is increased [88], but it is more commonly impaired in late-stage disease. In a study involving 98 HF patients referred for heart transplantation, an abnormal RV FWS was the strongest predictor of the combined outcome of HF hospitalisation, cardiovascular (CV) death, insertion of intra-aortic balloon pump or ventricular assist device, and need for heart transplant [89]. RV FWS was an even stronger predictor than LV GLS [89]. Multiple studies have confirmed RV FWS strain as an independent predictor of overall and CV mortality in patients with HFrEF-superior to TAPSE, FAC, and RV S' [90][91][92]. The sensitivity of RV strain to predict poor outcomes in HFrEF may be explained by RV FWS being the most precise functional measure that correlates with the degree of myocardial fibrosis [93].
Tricuspid regurgitation is most commonly secondary to pulmonary hypertension and/or annular dilatation in the setting of RV volume/pressure overload, usually from left heart disease [94]. "Primary" TR due to valve disease being uncommon-perhaps the most frequent manifestation being related to pacing leads. Accurate assessment of RV function is pivotal in decision-making about tricuspid regurgitation. In primary TR, delayed intervention on the tricuspid valve may lead to permanent RV dysfunction [95][96][97]. Impaired RV FWS in patients with functional tricuspid regurgitation (TR) is more sensitive at detecting RV impairment and associated with worse outcomes when compared to conventional measurements of RV function [98]. Notably, in this population, a significant proportion of patients had abnormal RV FWS but preserved FAC and TAPSE suggesting that myocardial dysfunction occurs prior to the fall in RV ejection fraction [98]. RV strain is a highly sensitive and prognostic tool in patients with tricuspid regurgitation. Future clinical applications to utilise strain to improve timing of tricuspid valve intervention is promising but clinical benefit remains to be determined [99].

Rationale for the Assessment of RA Function
The right atrium (RA) has often been neglected, undervalued, and poorly understood in its contribution to overall cardiac function. In consequence, assessment of RA function has only been assessed in a few disease states. Similar to the LA, the RA has phasic functions which can be evaluated by echocardiography and CMR [100].
Evaluations of RA size and function are valuable in the outcomes of at-risk conditions such as pulmonary hypertension, heart failure, atrial arrhythmias, and post-RV infarction. In a meta-analysis of 12 studies, including 1085 patients with PH, those with an enlarged RA area were associated with a poor prognosis [101]. Impaired RA reservoir and conduit function and impaired RA strain in patients with pulmonary hypertension also correlated with a worse functional class [102]. In patients with HFpEF, impaired CMRderived RA conduit and reservoir function were independent predictors of mortality [100]. Atrial arrhythmias are associated with RA dilatation in both 2D and 3D echocardiography [103]. Assessment of RA volumes has also demonstrated the ability of the RA to reverse remodel post radiofrequency ablation for AF [104]. Finally, RA function as assessed by reservoir and conduit RA strain demonstrated a prognostic role in patients post RV infarction [105].

Technical Aspects of RA Strain Measurement
Accurate assessment of RA function continues to be a challenge due to the impact of loading and off-axis imaging with echocardiography and the challenges of regional assessment with CMR [106]. Traditional 2D echocardiographic assessment of RA area from planimetry is susceptible to errors arising from geometric assumptions. RA volume assessment is not routine or supported by current guidelines due to the lack of standardised RA volume data [107]. RA strain may be assessed via both CMR and STE. CMR feature tracking assessment of RA strain has strong inter-and intra-observer reliability with good consistency with STE RA strain, however again the limitations of CMR revolve around access and cost [108]. STE RA strain aims to add to the assessment of RA function given the promising outcomes in RV, LA, and LV strains. However, there is still a significant gap in the literature of strong prognostic and diagnostic implications of STE RA strain. A review of the literature by Khan et al. in 2018 showed that only 59 studies were relevant to a PubMed search on STE RA strain. Of these, the majority (80%) were retrospective single studies highlighting the need for future prospective and randomised controlled trials to strengthen the clinical utility of STE RA strain [109].

Clinical Considerations
Unfortunately due to the lack of established normal values, the clinical utility of RA strain remains limited [110]. The largest published meta-analysis of RA strain involving 4,111 normal adults concluded there is limited clinical utility of RA strain due to the extensive variability in currently reported normal values of RA strain [111•]. Until technical capabilities improve, right atrial strain should continue predominantly as a research tool.

Conclusion
Significant progress has been made over the past few years in improving the diagnostic utility and furthering clinical applications of strain. Expanding beyond LV GLS, chamber-specific strain software and taskforce consensus statements have helped overcome the previous limitations of reproducibility and inter-vendor variability-most notably evident in left atrial strain. Further research should focus on confirming a consensus for assessment by 3-or 6-segment RV strain and eventual trials on clinical outcomes from strain-directed therapy.
Funding Open Access funding enabled and organized by CAUL and its Member Institutions Funded in part by an Investigator grant (2008129) from the National Health and Medical Research Council, Canberra, Australia.

Conflict of Interest
The authors declare no competing interests.

Human and Animal Rights and Informed Consent
This article does not contain any studies with human or animal subjects performed by any of the authors.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.