1/ #GLS is not an objective measure, it's totally method dependent, and therefore with no gold standard, and no possibility of validating measurements. Why is this?
2/ Let's go into the definition of strain. The Lagrangian definition is S = (L-L0)/L0, change in length divided by original length. For GLS, that means (roughly) longitudinal shortening / end diastolic length.
3/ Since longitudinal shortening can be measured by longitudinal M-mode as MAPSE, this means GLS can be measured as MAPSE / end diastolic length.
4/ In the HUNT study, Mean MAPSE was 1.58 cm, with less than 1 mm Difference between mean of 2, 4 or six walls. pubmed.ncbi.nlm.nih.gov/29399886/ Mean end diastolic mid LV length was 9.24 cm (unpublished), giving a mean GLS of -17.1%
5/But this measure is only related to this specific choice of reference length (denominator). We previously chose the straight line from apex to the mitral points, calculating strain per wall and mean. pubmed.ncbi.nlm.nih.gov/29399886/
6/ This was more robust, the straight lines were closer to wall length, giving a measure closer to wall strain. But it's evident that this mean WL is a little longer than LVL,
9,47cm, and mean GLS by this denominator was -16.3%, lower absolute, because of the higher denominator
7/ Following the curved wall, would give even longer WL, and lower GLS, evident from the fig. We didn't do this exercise, as manual drawing would be to variable, and HUNT3 is vivid7 data. It could be done now, by automated methods in HUNT4, with better data, (If interesting).
8/ So even with these straightforward methods, the choices of denominator influences strain values, and there is no ground truth. Speckle tracking opens a new can of worms, as the algorithms are "black boxes", proprietary to vendors, and subject to change w/software versions
9/ There are, however some general priciples. In general, ST GLS tends to give higher absolute values, around -19 - -20%. Thus, even having curved ROIs following the walls ST draws in the opposite direction from the method outlined in tweet 7/. Why is this?
10/ Speckle tracking in general not only have curved ROIs, but also tracks crosswise motion of the speckles. As the wall thickens, this means that speckles move inwards in the cavity, in this example it's the endocardial boundary
11/ This is most pronounced at the endocardial border, leads so, but still present in the midway, and least at the external LV border. Most applications use either endocardial border, or a thick ROI, where mean motion most closely corresponds to the midwall
12/ But what happens when tracking a curved boundary that moves inward toward the curvature centre? Exactly, it becomes shorter. This effect is there, even when there is no longitudinal shortening, best illustrated with circumferential shortening.
13/ Circumferential strain is negative (shortening), but as seen by the unmoving diameter, this is *only* due to inward motion, which is a function of external circumferential shortening *and* wall thickening. pubmed.ncbi.nlm.nih.gov/31673384/
14/ But this means that tracking derived GLS actually over estimates the longitudinal shortening, by incorporating some curvature shortening, which again is mainly wall thickening. In my opinion, this is another systematic error in ST derived GLS.
15/ And as this inward motion due to wall thickening is most pronounced at the endocardium (due to full wall thickening, as opposed to midwall, which only relates to outer half thickening), this is probably some of the basis for the unsound notion of "layer strain".
16/ In addition, the black box ST applications all have complex algorithms with different choices for
-Assumptions of LV shape and ROI width
-Number, size and stability of speckles
-Spline smoothing along the ROI and weighting of the AV -plane motion
-Etc.
17/ Interestingly, we developed an in-house application for strain, tracking kernels longitudinally by TDI and transversely by ST, calculating segmental and global strain. It gave nearly the same GLS as the linear method in 5/. pubmed.ncbi.nlm.nih.gov/19946115/ pubmed.ncbi.nlm.nih.gov/29399886/
18/ I would expect it to be subject to the same error by inward tracking as ST, but as this used TDI data with a low underlying B-mode FR and lateral resolution, the lateral tracking may have been so poor, the technical shortcoming offset the systematic error.🤯
🧵 on atrial systole. 1/ Already in 2001, did we show that both the early and late filling phase was sequential deformation propagating from the base to the apex. pubmed.ncbi.nlm.nih.gov/11287889/
2/ This means, both phases consist of a wall elongation wave, generating an AV-plane motion away from the apex. So what are the differences?
3/ Only e’ correlates with MAPSE, so the elastic recoil is finished in early systole, while a’ do not, so atrial systole is a new event, caused by the next atrial contraction. pubmed.ncbi.nlm.nih.gov/37395325/
🧵1/ Sorry, I accidentally deleted the first tweet in this thread, here is a new and slightly improved version. Looking at the physiology of AVC propagation velocity, there are confounders galore, so taking it as a marker of fibrosis, is premature, to put it mildly.
2/ Firstly, The AVC is an event of onset of IVR, i.e at a part of heart cycle with relatively high cavitary and myocardial pressure. This may contribute to wall stiffness, which again may affect (probably increase) wave propagation velocity.
3/ Secondly, This may affect AS patients; who may have a higher wall/cavity pressure at end systole than controls, and thus higher pressure related stiffness.
🧵1/ Looking at the physiology of AVC propagation velocity, there are confounders galore, so taking it as a marker of fibrosis, is premature, to put it mildly.
2/ Firstly, The AVC is an event of onset of IVR, i.e at a part of heart cycle with relatively high cavitary and myocardial pressure. This may contribute to wall stiffness, which again may affect wave prpagation velocity.
3/ Secondly, AS patients may have a higher wall/cavity pressure at end systole than controls, and thus higher pressure related stiffness.
🧵 On early diastole. 1/ It is important to differentiate relaxation and myocyte elongation. Relaxation means tension devolution, due to the removal of Ca, and dissolution of actin/myosin cross bridges. Elongation means volume expansion. They are not simultaneous.
2/ Myoccyte relaxation actually starts during ejection at the time of peak pressure, the decreasing pressure during ejection shows decreasing myocyte tension. pubmed.ncbi.nlm.nih.gov/6227428/
3/ Simultaneously, ejection continues, chiefly due to inertia, until overcome by the Ao-LV pressure gradient, when AV closes. Thus, there is simultaneous myocyte relaxation (tension↓) and volume ↓ (= myocyte shortening). Here is blood flow / myocardial deformation interaction
🧵1/ The E/A fusion in mitral flow with higher HR is well known, normally occurring around HR 100.
2/ also, it should be well known that this occurs because the diastole shortens more with high HR than systole. But why?
3/ In an early study of intervals during exercise, we showed that the RR-interval and DFP, but not LVET shortened in parallel < HR 100. > HR 100 (< RR 600) Both LVET, DFP and RR interval shortened in paralell, but at a slower rate. pubmed.ncbi.nlm.nih.gov/14611824/
🧵 As for MAPSE, we showed in HUNT3 thatpwTDI S' varies between mitral ring sites. LV global S' must be averaged, but we have shown that the difference between mean of septal/lateral and of septal/anterior/lateral/inferior is negligible.
2/ Values are age dependent, and in fact mean of 2 walls was 8.37 cm/s, and of four walls 8.4 cm/s, the difference was statistically significant, but totally un interesting as lower measurement limit of pwTDI is 0.1 cm/s. folk.ntnu.no/stoylen/strain…
3/ But what about diastolic velocities? variation of e' between sites is present as for S', as shown previously in HUNT3. It is common to average lateral/septal, but I haven't found any comparisons between two and four sites, so I looked at that in HUNT3 and found: