Exercise cardiovascular magnetic resonance: development, current utility and future applications

Directly analogous to treadmill stress echocardiography, treadmill Ex-CMR is performed to achieve the required exercise intensity/ target heart rate (THR). The patient is then rapidly transferred into the CMR scanner for post stress imaging. Treadmill Ex-CMR has progressed from exercising outside the scanner room [33], to the development of a CMR compatible treadmill to allow exercise to take place inside the scanner room [34] and eventually performed adjacent to the CMR scanner [27] (Fig. 1). For ischaemia studies, this progression has reduced the ‘cooling off period’ from peak stress to image acquisition, as even a 60–90 s delay in performing stress echocardiography image acquisition, has demonstrated recovery of ischaemic regional wall abnormalities and thus decreases the sensitivity of ischaemia detection [35, 36]. Numerous studies in exercise echocardiography have demonstrated the differences between peak and post stress imaging, specifically demonstrating that peak stress imaging has superior sensitivity and accuracy at detecting ischaemia than post stress imaging [35, 37‐40]. A direct ‘head-to-head’ comparison in stress echocardiography demonstrated that peak stress supine bicycle echocardiography was superior to post stress treadmill echocardiography in ischaemia detection [41]. Consequently, stress echocardiography guidelines recommend post stress imaging be accomplished in under 60 s [26]. However, as CMR can detect ischaemia by assessing myocardial perfusion in addition to assessing wall motion abnormalities, treadmill Ex-CMR may be less time sensitive post exercise cessation than treadmill echocardiography. Varying transfer times have been achieved in treadmill Ex-CMR studies (Table 2) [27, 31, 33, 34, 42‐45]. Since progression to a CMR compatible treadmill adjacent to the scanner, all studies demonstrate scan initiation in under 30 s, as shown in Video 1 (with the exception of La Fountain et al. in which removal of the face mask assessing oxygen uptake prolonged transfer time [44]) and cine imaging completion in under 60 s [42, 43, 45].

In-scanner Ex-CMR can be performed by supine cycle or stepper ergometer, upright cycling in an open magnet, exercise in the prone position or using isometric handgrip; each method is reviewed in Table 1. In-scanner exercise overcomes the main limitation of treadmill Ex-CMR of heart rate reductions between exercise cessation and image acquisition. Imaging during exercise does however have difficulties. Exercise invariably creates movement, increased respiratory motion and interference to the surface ECG, all of which increase with increasing workload. Movement can be reduced with the use of straps or harnesses, but not entirely, especially at higher levels of exercise. Breath held images can be performed during exercise but are non-physiological and difficult at higher exercise intensities [29]. Imaging during free breathing can cause significant though plane motion, making obtaining reliable flow measurements difficult, with the pulmonary trunk especially challenging due to its short length before bifurcation [55]. ECG interference during exercise can create ghosting artefacts with gated images and as previously described, accurate 12 lead ECG monitoring with ST segment analysis cannot be performed during in-scanner exercise [29, 30, 52]. Finally, reaching maximal heart rate is more difficult with supine exercise compared with upright exercise; this is well documented in stress echocardiography with comparisons between treadmill and supine cycle exercise [41, 48, 56, 57]. One explanation is early termination due to leg fatigue [41, 53], thus VO_2max is often 10–20% lower in supine cycle exercise than treadmill exercise. Despite this, evidence from stress echocardiography demonstrates equivalency or superiority in detecting ischaemia over post stress treadmill exercise [41, 48, 56]. Indeed, comparatively higher blood pressure attained during supine ergometry [41, 48, 57, 58], results in a similar rate-pressure product to treadmill exercise, such that THR during supine exercise are generally lower when compared to the same exercise intensity in the upright position. Despite the described difficulties of performing in-scanner exercise CMR, techniques have been adapted, with the use of the supine cycle ergometer, such that it is possible to perform in-scanner Ex-CMR to maximal intensity heart rates with imaging during exercise to assess either bi-ventricular function or great vessel flow [29, 55], but often not both, due to time constraints of scanning at incremental levels during or immediately post exercise.

The first published use of a CMR compatible cycle ergometer was in 1995 using a 0.5 T whole body scanner to measure exercise changes in aortic flow [59]. Studies utilising commercially produced cycle ergometers followed in 1998 with the use of the Lode BV MR compatible ergometer (Fig. 3) on a 1.5 T CMR scanner [60]. Whilst the majority of Ex-CMR cycle ergometer studies use this system [29, 54, 55, 60‐80], some institutions have created custom made CMR compatible cycle ergometers [25, 81, 82]. Other approaches include the supine CMR compatible ‘stepper’ ergometer, that utilises an up/down motion, such as the Lode BV up/down ergometer [83‐85], Ergospect cardio-stepper [86] and custom built supine steppers as demonstrated in Fig. 4 [87]. Studies using stepper systems report the benefit of reduced upper body motion, thus reduced motion artefact, and less restriction of leg movement compared with cycle ergometers. However, the up/down motion recruits less muscle mass than the cyclical motion. Thus no study has demonstrated supine ‘stepper’ ergometer Ex-CMR to maximal intensity, as has been demonstrated with supine cycle ergometers (Video 2) [29].

Cheng et al. demonstrated the feasibility of assessing pulmonary artery flow during continuous exercise to moderate intensity in adults and children in a 0.5 T vertical open bore scanner [101]. Although upright cycling may be more tolerated than supine exercise, it requires the use of an open low field CMR scanner, with benefits of easing claustrophobia, but inherent issues of lower SNR. CMR is feasible at lower field strengths [102], however, although the scanners are commercially available they are not in mainstream use, as such very few Ex-CMR studies have utilised this modality.

Isometric exercise involves the contraction of skeletal muscle without the elongation of the muscle, as such is also called static exercise [103]. This is feasible during CMR by IHG or isometric bicep exercise [104]. IHG exercise comprises the constant squeezing of a lever on a hand dynamometer, generally to a percentage of the subjects maximum force. The technique only allows for modest increases in heart rate, typically 10-20 bpm above resting rates, but causes minimal movement. As such the technique has mainly been used for CMR spectroscopy (MRS) or coronary artery flow imaging where minimal movement artefact is pivotal and minimal heart rate increases are acceptable. Weiss et al. performed the seminal work with IHG-Ex CMR, developing the ³¹P CMR spectroscopy (CMRS) stress test which remarkably can detect ischaemia in patients with CAD, despite minimal stress and increases in heart rate [105].

Prone Ex-CMR was first employed by Conway et al. who performed Ex-CMR studies using knee extension with a custom system of straps, cables, pulleys and weights [106]. Numerous subsequent Ex-CMR studies have similarly used prone Ex-CMR performing alternative knee flexion with ankle weights [107‐109]. Low-moderate intensity exercise is feasible by this approach, with the most significant response from the custom knee extension system by Conway et al., with a mean stress heart rate of 119 bpm. This technique has other inherent limitations; exercising whilst prone is an unnatural form of exercise which uses weights or resistance bands attached to the legs, increasing the resistance can be labour intensive, requiring alterations during exercise/scanning or exercise cessation. Prone Ex-CMR often requires an auditory cue from a metronome to determine work speed, however if this isn’t strictly adhered to, then the exact workload is unknown. As such, only Conway et al. employed incremental resistance by increasing the attached weights to the pulley system used. As such, prone Ex-CMR is not ideal to assess incremental levels of exercise intensity or where strict heart rate increases or levels are required.

Numerous studies have assessed regional wall motion and/or myocardial perfusion with Ex-CMR, all utilising treadmill exercise with post-stress imaging to achieve this (Table 3) [27, 31, 33, 34, 42, 44, 45]. Supine cycle ergometer myocardial stress perfusion has been demonstrated as feasible in healthy subjects [54] but has not yet been utilised to assess patients with CAD. The original treadmill work assessed 27 patients referred for coronary angiography with treadmill stress CMR and demonstrated a sensitivity and specificity of 79 and 85% respectively, for detecting stenosis (> 70%) [33]. The Ohio State University research group developed a CMR compatible treadmill [34] and utilised this in the EXACT trial [45]. The EXACT trial (exercise CMR’s accuracy for cardiovascular stress testing) is a multicentre prospective observational study. Four participating centres recruited 210 patients who were clinically referred for treadmill radionuclide single photon emission computed tomography (SPECT) and directly compared this with treadmill stress CMR. Patients had a rest ^99mTc SPECT rest scan followed by resting CMR cine images. An in room treadmill stress test was then performed with ^99mTc tracer injected at peak stress. Patients were then rapidly transferred to the CMR scanner for CMR stress images, including cine and myocardial perfusion imaging. Stress SPECT images were subsequently performed in an adjacent gamma camera. All images were acquired in the same day from the same treadmill stress. Ninety-four of the patients had coronary angiography (X-ray or CT coronary angiography). This multi-centre trial demonstrated superiority of Ex-CMR over exercise SPECT for identifying > 70% at coronary angiography. Sensitivity, specificity, positive and negative predictive values for Ex-CMR were 78.6, 98.7, 91.7 and 96.3% respectively, compared to 50, 93.7, 58.3 and 91.5% respectively for exercise SPECT. The EXACT trial was the first, and to date the only, multicentre Ex-CMR trial and demonstrated the diagnostic capabilities and clinical utility of treadmill Ex-CMR in CAD. The initial and downstream costs of treadmill Ex-CMR have being assessed by the recently completed and under review, multicentre EXACT-COST trial, which randomised patients to either treadmill stress SPECT or treadmill Ex-CMR. Treadmill Ex-CMR has similarly shown a promising comparison with exercise echocardiography in a pilot study of 28 patients, with Ex-CMR showing more myocardial segments being adequately visualised than by exercise echo [42]. In addition to CAD diagnosis, treadmill stress Ex-CMR has potential prognostic utility in predicting adverse events in those with known or suspected CAD. In a study of 115 patients with an indication for stress imaging, wall motion abnormalities were assessed at rest and post treadmill stress via Ex-CMR. The presence of stress induced regional wall motion abnormalities on treadmill Ex-CMR better identified those at risk of future events (death, myocardial infarction & unstable angina prompting admission) when compared to exercise ECG alone [43].

A recent supine cycle Ex-CMR feasibility study demonstrated post exercise T1 mapping could be a surrogate marker in detecting myocardial blood flow. In 14 CAD patients, post exercise T1 reactivity was associated with the severity of myocardial perfusion defect on MPS-SPECT. Therefore, Ex-CMR T1 mapping shows potential for detecting myocardial ischaemia [110].

Ex-CMR studies using IHG stress can reproducibly assess coronary artery endothelial function by assessing coronary artery cross-sectional area change and coronary flow by velocity encoded CMR [111]. Studies have demonstrated endothelial dysfunction in CAD in the form of paradoxical vasoconstriction with exercise [112], assessed the reproducibility of this technique [113] and shown that the normal vasodilatory response of healthy coronary arteries is mediated by nitric oxide [114]. Coronary artery blood flow assessment after supine bicycle exercise is also feasible, correlating well with invasive measures [115], and has been used to demonstrate, in a randomised double-blinded crossover trial, that hormone replacement therapy and high dose statins increase coronary artery blood flow in postmenopausal patients without CAD [116]. Thus, Ex-CMR may have a future role in research studies investigating CAD pathogenesis in low risk populations.

Ex-CMR studies show promise in providing an additional tool to differentiate the athletic heart adaptation from cardiomyopathy, and for risk stratifying endurance athletes against RV arrhythmias. Ex-CMR to maximal intensity on 14 endurance athletes after a 150 km cycle race, showed that the acute RV dilatation and reduced RV ejection fraction (RVEF), previously demonstrated with endurance exercise, worsened during further exercise and was not the result of confounding post-race variables such as increased afterload and autonomic activation [75]. Using Ex-CMR, endurance athletes with RV arrhythmias demonstrated more RV dysfunction during exercise, despite normal resting RV function, than athletes without RV arrhythmias and healthy controls [65]. In terms of differentiating the athletic heart from cardiomyopathy, when 10 healthy endurance athletes were compared with 9 endurance athletes with subepicardial fibrosis and 5 patients with dilated cardiomyopathy, they demonstrated a significantly greater augmentation of LV ejection fraction (LVEF) during exercise of 14 ± 3%, compared with 4 ± 3% and 5 ± 6% respectively. The endurance athletes with subepicardial fibrosis had similar resting haemodynamics and exercise capacity to the healthy endurance athletes, suggesting that Ex-CMR may be helpful in differentiating healthy endurance athletes from those with fibrosis [63].

CMR has significant benefits over alternative non-invasive imaging modalities for the assessment of congenital heart disease, especially for complex lesions and right heart pathologies. These benefits have been utilised with supine ergometer Ex-CMR to assess the physiological responses to exercise in a variety of conditions (Table 4).

Despite normal resting biventricular function, patients with prior atrial correction for transposition of the great arteries (TGA) demonstrated little or no increase in EF & stroke volume (SV) on Ex-CMR compared with healthy controls. The abnormal RV response to exercise correlated with exercise intolerance in these patients [71]. A subsequent study comparing haemodynamic response between Ex-CMR and dobutamine stress CMR also demonstrated no increase in SV and EF with exercise in atrial-corrected TGA patients, but a significant increase of SV and EF in controls during exercise and both groups during dobutamine stress CMR. This study demonstrated the two stress techniques could not be used interchangeably in this cohort of patients and re-affirmed the abnormal exercise response in atrial-corrected TGA patients. The differences in stress response between dobutamine and exercise are likely due to the differing effects on pre and afterload and could be present in other cardiac disease processes. Therefore, the type of stressor must be taken into account when interpreting the different methods [67]. A subsequent Ex-CMR study demonstrated a divergent cardiac response during Ex-CMR between atrial corrected TGA and congenitally corrected TGA (CC-TGA) patients, demonstrating the need for caution in performing pooled analysis of systemic RV (s-RV) patients. At rest, atrial corrected TGA patients had significantly higher global circumferential strain but similar global longitudinal strain and s-RVEDVi and ejection fraction. CC-TGA patients demonstrated more pronounced septal interstitial expansion. During Ex-CMR, atrial corrected TGA patients demonstrated deteriorating s-RVEDVi and RV SV during exercise compared with an unchanged s-RVEDVi and augmented RV SV in CC-TGA patients [80].

Ex-CMR has been important in demonstrating causes of exercise intolerance in patients with Fontan circulation. Pedersen et al. investigated patients with total cavopulmonary connection (TCPC), a palliative Fontan type operation, using supine bicycle Ex-CMR. At rest blood flow was lower to the smaller left lung. During exercise the flow increased proportionally with systemic venous return, IVC flow doubled and SVC flow increased slightly. The ratio of flow distribution to the lungs was unchanged with exercise, contrary to earlier resting studies demonstrating preferential streaming of IVC flow into the left pulmonary artery (LPA) and SVC to the right pulmonary artery (RPA) [122, 123]. The study concluded that pulmonary vascular resistance, rather than anastomosies geometry, was the major factor effecting exercise flow distribution and thus exercises intolerance [98]. This assertion was further demonstrated in an Ex-CMR study in 10 Fontan patients, in which rest and exercise systemic volumes were assessed with simultaneous invasive radial artery and pulmonary artery pressure measurements before and after sildenafil. Sildenafil improved cardiac index during exercise with a decrease in total pulmonary resistance index and an increase in stroke volume index [78]. The ‘respiratory pump’ theorized in the Fontan circulation has been demonstrated in 10 Fontan patients who showed increased ventricular filling during inspiration which persisted during Ex-CMR. This may provide a rationale for respiratory muscle training in this patient group [79]. Subsequently, Wei et al. studied the significance of varying respiratory effects on flow in the TCPC anatomy by using real time phase contrast imaging to assess the SVC, IVC, ascending and descending aortic flow during breath holding, free breathing and exercise conditions. The study demonstrated that whilst respiration affects IVC and SVC waveform, it did not significantly affect mean flow. The study therefore validated using breath held images, which benefit from reduced image artefacts, for routine mean flow assessment in TCPC anatomy patients. Additionally, the study demonstrated interchangeable mean flows between the IVC and descending aorta, justifying the use of descending aorta flow as a surrogate for IVC flow, which is difficult to assess during exercise due to diaphragmatic motion, in TCPC anatomy patients [97]. Khiabani et al. demonstrated the association of exercise performance with power loss in TCPC anatomies when performing an Ex-CMR study in 30 patients with Fontan circulation. Power loss calculations made from computational fluid dynamic simulations were performed on the in vivo TCPC anatomies and acquired flows, demonstrating that as cardiac output increased during exercise, power loss increased exponentially [119]. A subsequent Ex-CMR study demonstrated an inverse correlation between the TCPC diameter index, which accounts for vessel narrowing in the TCPC, and indexed power loss during Ex-CMR to the ventillatory anaerobic threshold, thus suggesting that reducing vessel narrowing and elevated power loss could improve exercise capacity and quality of life in patients with TCPC [120]. The diminished heart rate reserve often seen in Fontan patients was investigated in a recent Ex-CMR study and demonstrated as a likely secondary phenomenon rather than due to sinus atrial node dysfunction. 10 Fontan patients demonstrated an increased chronotropic response to exercise but an early plateau in cardiac output, compared with 20 healthy controls, caused by premature reductions in ventricular filling and stroke volume. This Ex-CMR study thus determined that the diminished heart rate reserve observed in Fontan patients was a result of abnormal cardiac filling rather than sinus atrial node dysfunction causing chronotropic incompetence [77].

Patients with prior surgically repaired VSD commonly have reduced functional capacity and often demonstrate resting retrograde ascending aortic flow. A recent Ex-CMR study demonstrated a decreased cardiac index in young adults exercised to submaximal intensity. This was determined secondary to a previously unrecognised increase in retrograde pulmonary flow with exercise along with previously demonstrated chronotropic incompetence in this patient group [61].

Ex-CMR has been demonstrated as feasible in repaired ToF in both adults [121] and children [85]. The seminal Ex-CMR study in repaired ToF patients by Roest et al. demonstrated an abnormal RV response to exercise compared with controls, but a decrease in pulmonary regurgitation [117]. A more recent study demonstrated reduced biventricular contractile reserve and aortic distensibility in repaired ToF patients compared to healthy subjects during Ex-CMR, which may be an early sign of increased aortic rigidity [121].

An Ex-CMR study assessing the exercise biventricular response to percutaneous pulmonary valve implantation (PPVI) in patients with either pulmonary regurgitation or pulmonary stenosis of heterogenous congenital aetiologies, demonstrated that PPVI resulted in restoration of RVEF exercise reserves in pulmonary stenosis patients but only a mild augmentation of exercise SV post PPVI in pulmonary regurgitation patients [118].

Idiopathic PAH patients have shown worsened RV function and a failure to augment bi-ventricular stroke volumes during Ex-CMR when compared with healthy subjects [64]. A recent study found similar differences between healthy controls and PAH patients, but additionally, when healthy controls were exercised during hypoxia (breathing 12% oxygen), which is associated with transient acute pulmonary hypertension, this resulted in decreased RV contractile reserve, but not as dramatically as PAH patients exercising during normoxia [100]. A further increase in pulmonary artery pressure during exercise in idiopathic PAH patients is the presumed cause of worsened RV contractile reserve; this was demonstrated in an Ex-CMR study of 15 patients with chronic thromboembolic pulmonary hypertension (CTEPH), 7 patients post pulmonary endartectomy (PEA) and 14 healthy controls. Despite a normal resting RVEF and mean pulmonary artery pressure post PEA, the RVEF and pulmonary compliance decreased with exercise in the post-PEA group along the CTEPH group. This was in contrast to healthy subjects whose RVEF increased with exercise and pulmonary compliance only mildly decreased. Treatment with sildenafil in the post-PEA group did not alter resting RVEF or haemodynamics, but significantly decreased exercise mean pulmonary artery pressure and increased exercise RVEF. Thus this Ex-CMR study demonstrated that post-PEA patients display an abnormal pulmonary vascular reserve which can be partially reversed by Sildenafil [76].

Ex-CMR studies in diabetes mellitus have demonstrated that adolescents with type 1 diabetes mellitus (T1DM) have a reduced exercise capacity, and resting and exercise LV stroke volumes (which correlated to glycaemic control), compared with non-diabetic controls [88]. The same group demonstrated improved, but not normalised, LVEF in T1DM adolescents who undertook regular intense exercise for 20 weeks when compared with non-exercising T1DM controls, suggesting that the diastolic dysfunction could be partially reversed by regular intense exercise [89]. Impaired cardiac function during exercise was similarly demonstrated in adolescent type 2 diabetes mellitus (T2DM) patients when compared with obese non-diabetics and controls [81]. Similar to the studies in T1DM, the T2DM adolescents demonstrated smaller LV SV, EDV and ESV at rest and during exercise, with a smaller increase in EF than non-diabetic controls. As such, all 3 studies demonstrated a decreased cardiac reserve during exercise in adolescent diabetics, whether T1DM or T2DM, when compared with non-diabetic controls [81, 88, 89].

Ex-CMR studies in aortic regurgitation (AR) have demonstrated that isolated AR in children and adults decreased during ‘steady state submaximal exercise’ CMR, which equated to prolonged light in-scanner exercise on a custom built device [124]. Roberts et al. assessed the short term effects metoprolol and losartan had on exercise haemodynamics in chronic AR patients after supine exercise in a cross-over study, showing that with metoprolol there was a lower heart rate, greater AR regurgitant fraction, lower aortic distensibility and greater indexed LV EDV and ESV compared to Ex-CMR on losartan [125].

Weiss et al. performed the initial study in exercise CMRS, developing the ³¹P CMRS stress test by performing CMRS before, during and after isometric hand grip exercise in patients with CAD, non-ischaemic heart disease and healthy controls. The study demonstrated a deceased regional PCr/ATP ratio during exercise in patients with > 70% stenosis in the left anterior descending coronary artery (LAD) or left main coronary artery vs no change in healthy controls and patients with non-ischaemic heart disease. The decreased PCr/ATP ratio improved 2 min after cessation of exercise and did not recur in those that had subsequent revascularization [105]. The technique was further validated by Yabe et al. as a sensitive method to detect myocardial ischaemia in those with reversible defects on ²⁰¹T1 MPS-SPECT [127] and subsequently used to demonstrate decreased PCr/ATP ratios on exercise in cardiac transplant patients [128] and patients with Chagas heart disease [129].

Prone exercise ³¹P-CMRS has been used to investigate patients with T2DM and separately patients with hypertrophic cardiomyopathy (HCM). Resting ³¹P-CMRS studies have demonstrated decreased energy metabolism via a decreased PCr/ATP ratio in T2DM patients with functionally normal hearts [130]. A study by Levelt et al., employed prone knee flexion exercise stress to perform ³¹P-CMRS and an adenosine stress CMR to assess myocardial perfusion and oxygenation. The T2DM patients demonstrated a lower resting PCr/ATP ratio which decreased further with exercise compared to controls and a decrease in myocardial perfusion and oxygenation. The correlation between exercise myocardial perfusion and oxygenation and PCr/ATP ratio at exercise suggested that coronary microvascular dysfunction exacerbated the cardiac energetics during exercise [109].

Dass et al. used prone knee flexion exercise stress to perform ³¹P-CMRS and an adenosine stress CMR to assess myocardial perfusion and peak filling rates in patients with HCM and healthy controls. The HCM patients demonstrated a lower resting PCr/ATP ratio which decreased further with exercise compared to controls, a decrease in peak filling rates with exercise but no change in myocardial perfusion. PCr/ATP ratio and peak filling rates correlated strongly at rest and exercise in the HCM group, suggesting that abnormal cardiac energetics in HCM patients is a key mediator of exercise induced diastolic dysfunction commonly seen in this disease [108].