Limitations
The ability to distinguish partial volume border pixels and correctly determine whether pixels are contaminated by partial volume effects, or are true pathophysiology such as subendocardial or subepicardial fibrosis, is a residual and important issue. This issue is not only important in terms of the visual readout, but may introduce biases into quantitative measurements which become particularly significant in the context of more subtle diseases. One approach to reducing the bias in measurements is to restrict the measurement to the mid-wall region, although one must exercise caution for subjects with thin walled myocardium or patients with thin rims of subendocardial fibrosis.
The proposed approach to generating ECV maps resulted in excellent to good image quality maps in 83% of the cases. A number of factors contributed to cases of poor quality images. In some cases patient movement resulted in the slices being substantially different in appearance despite the fact that the same slice position was prescribed. Co-registration is by design only capable of correcting for in-plane motion and cannot compensate for through plane motion, although it may appear to partially compensate. In other cases, it is clear that the pre- and post-contrast images are at substantially different cardiac phases due to significant change in heart rate, despite adjusting the trigger window. Both of these factors might be improved at the time of acquisition by more careful inspection but this would complicate the workflow.
The interpretation probably can be improved by examining the pre-and post-contrast T1-maps for consistent appearance of position and cardiac phase as a quality control step. Furthermore, it is expected that the use of ECV maps will be in conjunction with pre-contrast T1-maps and conventional late enhancement images, and not read independently. As such, the ability to assess and quantify diffuse fibrosis in ECV images should provide increased diagnostic confidence since LGE image quality is generally excellent and will not share the same artifacts. Finally, the use of residual error maps (Figure
7) offers a potential as another quality metric to increase the confidence for a given region. In other words, if the error map indicates that the motion correction is high quality (low error), then the ECV values are more accurate.
Partial volume effects affect the measurement of ECV. Partial volume effects are a result of limits in in-plane spatial resolution as well as slice thickness and are observed at tissue borders such as for pixels between blood and myocardial tissue. Partial volume effects also affect the measurement of fine structures such as MI contributing to a larger variability. These same factors are also a limitation in LGE. These effects are less of a factor for measurements of globally diffuse disease.
The issue of achieving a “dynamic steady state” using a bolus [
12,
13] rather than slow infusion [
3,
9] has been studied by several researchers, however, the accuracy of this assumption and the corresponding impact on quantitative ECV measurements is unknown. In particular, it has been noted that steady state may take longer to achieve in acute myocardial infarction [
16], therefore further work is required to investigate this limitation.
Despite these issues, ECV maps may be readily incorporated into the clinical workflow and may provide diagnostic information where other methods are limited. It is not intended to replace existing methods such as LGE which are excellent at depicting focal lesions, but rather to be used in concert with other techniques. When ECV is abnormally elevated, it may not be clear whether this is due to fibrosis or edema which may be either diffuse or focal. In such instances, pre-contrast T1 or T2 maps, in addition to patient history and contextual imaging clues like signs of heart failure, may help to differentiate these mechanisms.
The presence of intra-myocardial fat may affect the estimates for ECV and has not been studied. In cases of fat with low fat fraction, voxels with partial volume of fat and myocardial tissue may lead to errors in T1-mapping which assume a mono-exponential model for single T1-species. Furthermore, fat is not in fast chemical exchange with the Gadolinium contrast agent and will not experience the same T1 shortening. Multi-echo fat water separated inversion recovery may be used for late enhancement [
24,
25] but is challenging to incorporate into a T1-mapping sequences. Fat water separated imaging may also be used in the combined assessment to identify the presence of intra-myocardial fat.
In this study, the generation of ECV maps was performed off-line and retrospectively. In order to wholly automate the generation of ECV images on the scanner at the time of imaging, hematocrit would either have to be entered manually or automatically queried from a clinical database. These options are potentials for improvement which remain to be explored in order to make the generation of ECV images wholly automatic on the scanner as a part of the clinical work flow. While it is possible to compute a partition coefficient map without using the value of hematocrit [
26], the variability in hematocrit (18.3% to 65% in this series) makes the partition coefficient less useful for measuring subtle changes in ECV expected for diffuse fibrosis.
The method for automatic segmentation of the blood pool was effective for mid-ventricular short axis or long axis views acquired in this study. For volumetric coverage which would include apical slices with a reduced size of the region of blood, it might be necessary to adapt this method to use blood values acquired at other slices with the proviso that they must be acquired in close time proximity such that the blood T1 remains the same.
T1-mapping accuracy and precision
The accuracy and precision of T1-mapping using the MOLLI method is affected by a number of factors including the protocol, the sequence design, patient specific factors related to heart rate and blood flow, and system adjustments such as center frequency and shim. The MOLLI method for T1-mapping is widely use and a number of adaptions to the protocol have been introduced to shorten the acquisition time. Further optimization the sequence and of pre- and post-contrast T1-mapping protocols based on the MOLLI strategy is the subject of ongoing research.
A shortened acquisition acquiring 7 images in 9 heartbeats (known as ShMOLLI) [
27] may be used to further reduce the breath-hold duration and achieve a heart rate independent T1 measurement by discarding samples for tissue with longer T1. The 11 heart beat protocol used in this study achieves a reliable T1 measurement by using a longer recovery between inversions without eliminating data. At longer T1’s the 9 heart beat ShMOLLI protocol uses only 5 images for T1-fitting, whereas the proposed motion corrected 11 heart beat protocol uses all 8 acquired images. The increased number of images reduces the fitting error.
In terms of the sequence design, we have recently determined that imperfect inversion efficiency leads to an underestimation of T1, representing the largest source of systematic error. This error may be greatly reduced by means of an improved inversion pulse design as well as by applying an appropriate correction. These corrections were not available at the time of this study. As a result, the normal values for T1 are underestimated by as much as 5-10%. However the systematic bias in T1-measurement results in only a small overestimation of ECV, estimated on the order of 1% or less in ECV units. This does not alter the main conclusions of this present study on the utility of ECV maps.