Heart failure (HF) is a complex cardiovascular disorder characterized by impaired cardiac function and increased morbidity and mortality worldwide [1, 2, 3]. Accurate assessment of myocardial deformation in human subjects has emerged as a valuable tool in evaluating cardiac function and identifying early abnormalities in patients with HF [2, 4, 5, 6].

Myocardial strain assessment by cardiac magnetic resonance (CMR) imaging is a useful tool to measure global and regional myocardial function and deformation quantitatively in CMR imaging [5]. It has been shown to offer insights regarding the patient’s prognosis, both in acute and chronic HF [6, 7, 8, 9, 10]. Left ventricular deformation can be quantified in three dimensions: longitudinal strain, circumferential strain, and radial strain [11]. Among the various imaging techniques available, feature tracking (FT-MRI) strain analysis has gained significant attention due to its ability to quantify myocardial deformation using standard cine CMR sequences [5, 12]. After manual or automated segmentation of the myocardium in end-diastole, tracking of distinct pixel patterns (around 15 mm²) throughout the cardiac cycle for the entire myocardium reveals the myocardium deformation field from which the strain in three main axes are computed: longitudinal, radial, and circumferential strain [13]. Previous studies have demonstrated the potential of FT-MRI strain analysis in discriminating between healthy individuals and patients with HF [14, 15, 16]. This technique has been proven to be as reliable as acquisition-based strain-measurements. Post-processing solutions that allow for FT-MRI strain analysis based on standard steady-state free precession (SSFP) sequences are now offered by many vendors [17, 18]. However, universal standards for the interpretation of FT-MRI results do not currently exist, and different vendors and methods for analysis can significantly affect deformation values [18, 19, 20].

The cardiovascular angiographic analysis system magnetic resonance (Caas MR) Strain (Pie Medical Imaging) has recently been implemented in the IntelliSpace Portal Suite (Philips Medical Systems Nederland B.B., Best, The Netherlands) to assess global longitudinal (GLS), circumferential (GCS), and radial strain (GRS). It relies only on a two-chamber (2CH) and a four-chamber (4CH) view to calculate GLS and uses a full stack short axis sequence to assess GCS and GRS. However, normal values for these measurements have not yet been established.

Although previous publications have reported myocardial strain parameters using the present patient cohort, this study provides novel insights by establishing reference values for GLS, GCS and GRS using the Caas MR Strain software and reporting the first direct comparison with the widely used CVI42 platform. This study also proposes potentially clinically applicable cut-off values for pathological strain patterns in HF patients, aiming to support decision-making in routine practice. The aims of this study were threefold: to establish cutoff values for GLS, GCS, and GRS that distinguish healthy volunteers from HF patients, to evaluate the discriminatory power of these parameters in different HF subgroups, and to validate the comparability and agreement between software solutions from different vendors.

To the best of our knowledge, this study is among the first to demonstrate that calculation of GLS using only two long axis (LAX) acquisitions, 2CH and 4CH, is feasible and shows very good agreement with CVI42-based GLS assessment (r = 0.86, p 0.001). Strong correlations were observed between Caas MR Strain and CVI42 for GCS (r = 0.83, p 0.001) and GRS (r = 0.76, p 0.001), supporting the robustness of this software across multiple strain components. Our study thus highlights reliable comparability in regards to Circle’s CVI42, which in turn has been validated against other FT strain solutions [12]. This finding could prove beneficial to allow for better comparison of results from different sites and could potentially improve patient follow-up [18, 19].

Our findings suggest that FT-MRI strain analysis using Caas MR Strain may help differentiate between healthy individuals and patients with HF. In this study, we were able to propose cutoff values for GLS above –14.6%, for GCS above –15.3% and GRS below 23.1% to define pathological findings. Given the higher SD for GRS, the values seem less reliable. This is in line with previous findings suggesting increased reproducibility for GLS and GCS but less for global radial strain measurements [19, 25, 26].

A recent study from our research group identified a GLS cut-off point ($\ge$–15%) that showed potential for distinguishing HF patients from healthy controls using CVI42-based strain analysis [27]. Our results using CAAS MR Strain suggest a similar threshold. Furthermore, the method allows for some discrimination regarding the underlying HF subgroups. However, we were not able to show a clear discriminatory power for GLS, GCS, and GRS to differentiate between HFpEF, HFmrEF, and HFrEF. Similarly, we were not able to show the widely accepted significant decrease in GLS and GCS in HFpEF patients as compared to healthy adults, however, a trend towards decreased strain even in HFpEF patients is clearly visible in our cohort. This is possibly due to the relatively low number of patients in this study and somewhat contradictory to established diagnostic algorithms for HFpEF using different imaging modalities [2, 15, 28].

While in our study, the classification of HF was conducted based on an imaging perspective according to LVEF and FT-MRI strain parameters, the classified categories can generally be defined as different stages or severity of clinical HF, such as the classification systems used by the New York Heart Association (NYHA). In future studies, the combination of strain analysis with established clinical classifications may improve phenotyping and aid clinical decision-making.

The lower GLS observed in our study, compared to the literature, likely reflects both physiological and methodological factors. First, our control group was relatively older (mean age 63 $\pm$ 9 years). Previous CMR-FT studies have reported age-related variations, with progressive reductions in GLS with increased age [13, 29]. Second, gender stratification was limited in our cohort. Prior work has shown that females exhibit higher (more negative) GLS than males, with differences of approximately 1–2% [26, 29]. Since our group was composed of a balanced mix of both sexes, the overall values would be expected to be lower than those reported in female-only or sex-stratified cohorts. Third, strain was quantified using Caas MR Strain, which applies block-matching and Lagrangian strain computation. Inter-vendor comparisons have demonstrated systematic differences of 2–3% for GLS between commonly used platforms, mainly attributable to differences in contour definition, Region of Interest (ROI) tracking, and algorithm implementation [18, 19]. Therefore, age dependency, gender effects, and vendor-related variability could contribute to a composite deviation of 4–6% relative to reference values. This aligns with the data observed in our cohort and strongly suggests that the differences are methodological rather than pathological.

In patients with advanced HF, myocardial thinning and impaired wall motion may reduce the accuracy of endocardial and epicardial border definition, which could decrease the accuracy of FT-MRI strain measurements. While previous studies have shown good agreement between FT and reference techniques (tagging or Strain-Encoded (SENC)) [12, 19], strain accuracy can be reduced as ventricular remodeling increases [25, 26]. In our study, contours were carefully reviewed and manually corrected by two experienced readers (EACVI level III), but this variability remains an inherent limitation of FT-based analysis in abnormal myocardium.

A strength of our study is the reproducibility of myocardial strain assessment. The excellent inter-rater ICC values observed for GLS, GCS, and GRS reflect low ROI-dependent variability, thereby providing quantitative error estimates for accuracy-dependent changes. This methodological robustness further supports the assumption of a high reliability of strain values measured by this novel tool.

Our study only reports values for 1.5 T and for images acquired with the use of acceleration technique SENSE, which is widely adopted in clinical practice. Therefore, although our results are specific to 1.5 T scanners and acquisitions using SENSE acceleration, they may still be cautiously extrapolated to other settings, as previous studies have suggested minimal influence of field strength and acceleration techniques on strain values [19]. Since strain rate has been shown to be of incremental value in HF patients, future analysis should ideally include these time-dependent parameters [14]. Future directions of strain analysis hold the potential for including prospective cohorts for prognostic evaluation and assessing large populations using artificial intelligence [30, 31].

Segmental strain analysis was also performed and is shown in Fig. 2. Although our primary objective focused on global strain values, segmental strain may provide additional diagnostic information. However, segmental strain is generally associated with higher variability and lower reproducibility, which limits its immediate applicability in clinical practice. Future studies with larger cohorts should explore the diagnostic potential of segmental strain values in different HF populations.

This study has several limitations. We did not report age or gender-specific values, which could be beneficial since it has been shown that age is a potential confounder and younger patients are reported to have more pronounced myocardial deformation values [13, 29]. However, since we chose to study a group of HF patients, our values from relatively old volunteers (mean age 63 $\pm$ 9) are perhaps relevant to identify truly abnormal values among typical cardiac patients. A larger study population would allow reporting of age-specific normal values. The literature provides somewhat contradictory findings regarding gender, with most studies finding differences in strain values according to gender, although no uniform consensus is reported [13, 19, 26, 29]. The time required for analysis with both platforms was not systematically recorded; however, based on the authors’ experience, both tools provide semi-automated workflows with comparable post-processing time. Finally, the healthy control group was not age-matched to the HF cohort. This was due to the challenge of including older controls without cardiovascular comorbidities or subclinical abnormalities.

CMR imaging assessment of left ventricular myocardial strain using Caas MR Strain software reliably identifies HF patients. Discrimination between HF subgroups by GLS, GCS, and GRS showed statistically significant trends. However, given the limited sample size, these findings should be considered exploratory and require confirmation in larger studies. Inter-vendor agreement was most robust for GLS and GCS, but less robust for GRS. For practical clinical use, we propose cut-off values for GLS above –15%, GCS above –15%, and GRS below 23% to better define pathological changes.

Patients signed informed consent regarding publishing their data.

All data will be available from the corresponding author upon reasonable request.

MB, DH, H-DD, and SK conceived and designed the study. MB, DH, SK acquired clinical data. SC and KJW reviewed all scans and performed blinded strain analysis. KJW executed the statistical analysis. KJW and ICS drafted the manuscript. GV assisted with methodology and figure design. PD, KC, RT, SW, JV, REB, NM and CS revised and amended critical parts of the manuscript and approved the final version of this manuscript. All authors contributed to the interpretation of the data, revised and amended critical parts of the manuscript and approved the final version of this manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

The study protocol, inclusion of study participants, data access, and usage fully comply with the Declaration of Helsinki. All studies were reviewed and approved by the Charité–Universitätsmedizin Berlin Ethics Committee (Approval number: EA4/112/16) complied with the Declaration of Helsinki and was registered at the German Register for Clinical Studies (DRKS) (registration number: DRKS00015615). Informed consent was obtained from all individual participants included in the study.

We thank Corinna Else, Alireza Khasheei and Mandy Domning for their assistance.

The authors declare that this study received funding from Philips Healthcare, DZHK (German Centre for Cardiovascular Research), Partner Site Berlin, Germany and Myocardial Solutions. This work was partially funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)-SFB-1470-B06. The funders were not involved in the study design, collection, analysis, interpretation of data, and the writing of this article or the decision to submit it for publication. ICS received funding from the European Association of Cardiovascular Imaging (EACVI Training Grant 2025).

Sebastian Kelle reports grants and other support by the DZHK (German Center for Cardiovascular Research), Partner Site Berlin, Philips Healthcare, BioVentrix, Berlin-Chemie, Merck/Bayer, Novartis, AstraZeneca, Siemens and Myocardial Solutions outside of the submitted work. SK was also on the advisory board for Merck/Bayer, BioVentrix, and Myocardial Solutions. All other authors declare that they have no relationships relevant to the contents of this paper to disclose. Christian Stehning is an employee of Philips Healthcare or has a financial relationship with the company. Gaston Vogel is an employee of Pie Medical Imaging BV or has a financial relationship with the company. However, the company had no role in the handling or conduct of the study. The authors had full access to all data in the study and take full responsibility for the integrity of the data and the accuracy of the data analysis.