Retrospective temporal resolution interpolation alters myocardial strain quantification on compressed sensing cine CMR.

Compressed sensing (CS) is a rapidly developing technique that can acquire functional cines of the heart within seconds while free-breathing and it is ideal for assessing cardiac function in non-typical conditions such as when patients are sedated or anaesthetized or undergoing stress exams. CS cines can further include retrospective temporal resolution interpolation (INTP_TR) to improve the frame count per heartbeat, and the impact of INTP_TR on biventricular functional measurements is unknown. We investigated the impact of INTP_TR on left and right ventricular volumetry and strain measurements of CS cines. Nineteen patients with 51 different CS acquisitions were prospectively enrolled. CS cines were acquired at rest, under adenosine stress, oxygen inhalation or while under general anaesthesia with mechanical ventilation. From the same CS acquisition, a dataset with and without INTP_TR were generated by the scanner. The outputs were separated and analysed by blinded readers for left and right ventricular volumetry, as well as systolic and diastolic strain parameters using feature-tracking techniques. Measurements were compared between the INTP_TR and non-INTP_TR outputs. Similar measurements were obtained for biventricular volumes and ejection fraction independent of INTP_TR. Peak strain was significantly underestimated on INTP_TR cines for both longitudinal and circumferential orientations (p < 0.01). Nevertheless, good-to-excellent correlations were observed between the two measurements (r > 0.65, p < 0.01), and there was still a high area under the curve (AUC ≥ 0.95, p < 0.01) for detecting abnormal patients defined by strain analysis on the standard segmented cine. INTP_TR especially negatively influenced strain rates analysis, as many strain rate curves were deemed unusable with this technique. These findings were consistent independent if the patient was in a resting, stress or anaesthetized condition. Although INTP_TR is a feature which improves temporal resolution on CS cines, quantification of biventricular strain and strain rates is not feasible or comparable, thus, feature tracking analysis should be performed on non-INTP_TR data. However, volumetry and ejection fraction analysis are consistent independent of which output is analysed.