OMR overcomes most of the problems associated with echo, and does not involve the ionising radiation or need for contrast agents with EBCT. The free choice of imaging planes and excellent tissue visualisation mean that virtually all images are of sufficient quality for LV mass determination, and the technique has been well validated. The method involves imaging a stack of contiguous slices through the left ventricle, with the myocardial volume measured from each image slice summed to obtain the total myocardial volume, employing Simpson’s rule (figure 2.3). When multiplied by the density of myocardial tissue (1.05 g/cm®), the LV mass is obtained.
2.2.3.1 OMR technique used in this thesis
For the most accurate and reproducible measurements, the image stack should be parallel to the true LV short axis, which minimises errors due to partial volume effects of the myocardium within the image plane. The short axis was identified by first piloting a vertical long axis (VLA) plane from transaxial images, which passes through the centre of the mitral valve and apex of the LV. A horizontal long axis (HLA) plane was then obtained by imaging a plane perpendicular to the VLA, again passing through the centre of the mitral valve and apex. From the HLA, a stack of short-axis images was obtained, covering the length of the LV. ECG-gated cine-MR images were required in order to measure the LV mass at a single time point within the cardiac cycle (the standard is end-diastole - the frame immediately at the QRS of the ECG). Each cine image slice was acquired within a single breath-hold, removing motion artefact due to respiratory movement.
Studies were performed with a custom-built mobile cardiac magnetic resonance scanner (0.5 Tesla; imaging software; Surrey Medical Imaging Systems, Surrey, UK) belonging to Royal Brompton Hospital. The major advantage of this particular unit was its portability which allowed scanning of the subjects at their own base, minimising disruption to their usual schedule and maximising the uptake. It would not have been possible to include as many subjects as were obtained at the army training regiment in Bassingbourn, Hertfordshire (see chapter 3) without the presence of the scanner on site.
Dr. Myerson performed all the CMR scans, having been trained in the technique at the Royal Brompton Hospital, and conducted the image analysis with in-house software, CMRTools (©Imperial College), blinded to genotype and drug status, over a single time period.
Figure 2.3
Diagrammatic representation of a left ventricle with short axis imaging planes and typical CMR images obtained.
2.2.3.2 Accuracy and reproducibility
The accuracy of CMR measurements of LV mass has been yalidated using post mortem hearts, imaged in vivo in the case of animal studies 2 0 9 -2 1 4 ex-vivo (post
autopsy) for human hearts (table 2.1). These show excellent agreement between the OMR-obtained and true LV masses, with differences of 3-5% and standard deyiations of the order of 10g (95% Cl » 19g) in the canine studies and 8g (95% Cl « 15g) in the human ones. Giyen the 95% confidence interyals for echo-deriyed LV mass of 57-190g,^^®'^®® this represents a significant improyement in accuracy.
Table 2.1
Accuracy of CMR-determined LV mass in human and animal studies. Values are compared to post-mortem deriyed LV mass. SDD = standard deyiation of the difference between the two measurements; C.l.= confidence interyal (=1.96 x SDD); Ml = myocardial infarction; LVH = left yentricular hypertrophy. ^Standard error of the estimate from regression equation.
Study n SDD 95% G.I. Mean difference Mean % difference Notes Bottini Katz 6 10 8.9 g 7.4 g ± 17.5 g ± 14.5 g 0.7 g 10.2 g 4.0 % 5.3 % Human Human McDonald ^ 10 1.8 g ± 3.5 g 4.4 g 5.2 % Canine
Shapiro 10 6.7 g* ± 13.1 g Canine normal
8 8.7 g* ±17.1 g Canine post-MI
Caputo 13 13.7 g* ± 26.9 g 10.0% Canine normal + LVH
Keller 10 3.5 g ±6.9 g 6.8 g 13.3% Canine
Maddahi 8 4.9 g* ±9.6 g Canine; in vivo
9 3.4 g* ±6.7 g Canine; dead {in-situ)
Florentine 11 13.1 g* ± 25.7 g Canine + feline
Canine studies 79 7.0 g ±13.7 g
(Total & means)
Of greater importance for assessing changes in LV mass, both for indiyiduals and research studies, is the reproducibility of LV mass measurements by CMR. This encompasses inter-study (i.e. test-retest reliability) and inter- and intra-obseryer yariability in yalues. Again these are excellent for CMR, particularly for human studies (table 2.2). Inter-study yariability is reported at ~5% with a standard deyiation of the difference (SDD) of ~10g (95% Cl »19g). Intra and inter-obseryer yariability are also
good, being 6.1% and 5.1% (average) respectively. These values are equivalent, if not better, using the newer breath-hoid fast-acquisition technique used for this work.^^®
Table 2.2
Reproducibility of CMR-derived LV mass measurements from human studies. Values are standard deviations of the difference (SDD) between successive scans (g) or % variability. LVH = ieft ventricuiar hypertrophy; DCM = diiated cardiomyopathy; Ml = myocardial infarction.
Study n Inter-study Intra- observer
Inter observer
Notes
Bottini 4 8.2 g Normal subjects
Germain 20 11.2 g (6.7%) Normal subjects
Yamaoka 10 5.8 g 17.8 g Normals, LVH & DCM
Katz 10 6.1 % 7.2 %
Semelka 11 5.2 % 4.4 % Normal subjects
Semelka 11 4 .7 -6 .1 % 3.4 % DCM
8 3.5 - 4.8 % 5.5 % LVH
Matheijssen 8 3.6 % 3.6 % Ml
Beilenger 15 2.8 % 1.6% 2.4 % Normal subjects
15 3.0 % 2.7 % 3.1 % DCM
15 2.2 % 3.2 % LVH
Bogaert 12 4.4 % 4.1 % 4.2 % Normal subjects
Mean values 4.5 % (n=92) 3.4 % (n=75) 4.1 % (n=105) 2.2.3.3 Limitations of CMR
Patient factors can sometimes limit the usefulness of the technique. Due to the enciosed nature of the MR scanner, some people find this too claustrophobic though in practice, only two subjects found this to be a problem.
The same restrictions as for any MR scanner apply for patients with craniai aneurysm clips, ocular foreign bodies and pacemakers though none of these conditions occurred in the study groups.