CMR-Analysis (including machine learning)
Karl Jakob Weiss, MD
Physician
German Heart Center of the Charité
Berlin, Berlin, Germany
Karl Jakob Weiss, MD
Physician
German Heart Center of the Charité
Berlin, Berlin, Germany
Djawid Hashemi, MD
Physician / Clinician Scientist
Charité – Universitätsmedizin Berlin
Berlin, Berlin, Germany
Patrick Doeblin, MD
Cardiologist
German Heart Center Charité, Germany
Radu Tanacli, MD
Research Fellow
German Heart Center Berlin
Berlin, Berlin, Germany
Moritz Blum, MD
Resident
Charité – Universitätsmedizin Berlin, New York, Germany
Rebecca Beyer, MD
Physician
German Heart Center Berlin
Berlin, Berlin, Germany
Matthias Schneider, MD
Physician
Charité – Universitätsmedizin Berlin, Berlin, Germany
Hans-Dirk Duengen, MD
Physician
Charité – Universitätsmedizin Berlin, Berlin, Germany
Frank Edelmann, MD
Physician
Charité – Universitätsmedizin Berlin, Berlin, Germany
Burkert Pieske, MD
Head of Departement
Charité – Universitätsmedizin Berlin, Berlin, Germany
Sebastian Kelle, MD, FSCMR
Cardiologist
German Heart Center Berlin
Berlin, Berlin, Germany
Feature tracking magnetic resonance imaging (FT-MRI) myocardial strain reliably identifies heart failure (HF) patients with reduced (HFrEF), mildly reduced (HFmrEF), or preserved (HFpEF) left ventricular ejection fraction (LVEF).(1,2) Assessment of the global longitudinal strain (GLS), global circumferential strain (GCS), and radial short axis strain (GRS) are assessable by software analyses from multiple vendors. However, the values differ significantly.(3) Caas MR Strain - part of Pie Medical Imaging’s CMR software platform (Pie Medical Imaging BV, Maastricht, The Netherlands) has recently been implemented in the IntelliSpace Portal Suite (Philips Healthcare, Best, The Netherlands). Thus far, standard values for this software for different heart failure entities have not been established.
Methods:
At a 1.5 T Philips Achieva scanner, 40 patients underwent cardiac magnetic resonance (CMR) imaging, of which ten healthy controls and ten patients with HFrEF (defined as signs and symptoms of HF and a left ventricular ejection fraction (LVEF) of < 40%), HFmrEF (LVEF 40-49%), and HFpEF (LVEF ≥ 50%) each.(4) Analysis was performed using Caas MR Strain software. GLS, GCS, and GRS were quantified using 4-chamber-view, 2-chamber-view, and short-axis cine images. Statistical significance was assumed at p < 0.05.
Results:
Myocardial strain assessment by Caas MR Strain reliably discriminates between healthy volunteers and HF patients by GLS [median -15.5% (95% confidence interval (CI) -16.6%, -14.0%) versus -12.2% (CI -13.7%, -10.1%); p < 0.01 ], GCS [-15.9% (CI -17.7%, -14.4%) vs. -11.1% (CI -12.6%, -10.4%); p < 0.01] and GRS [26.5% (CI 22.5%, 29.2%) vs. 13.1% (CI 10.0%, 15.2%); p < 0.01]. GLS, GCS, and GRS differed among HF entities, significant differences are denoted in Figure 1 and Figure 2.
Conclusion:
FT-MRI myocardial strain analysis using Caas MR Strain software reliably identifies HF patients. We propose cut-off values for GLS above -14%, GCS above -13%, and GRS below 22% to define pathological findings. Discrimination between the different HF entities is potentially feasible by GLS, GCS, and GRS although greater patient numbers are needed. In the future, comparing these values to strain measurements from other vendors could further elucidate the clinical relevance of inter-vendor discrepancies in strain measurements.