Serial MRI studies over 12 months using manual and atlas-based region of interest in patients with amyotrophic lateral sclerosis

Amyotrophic lateral sclerosis (ALS) is characterized by the progressive loss of lower motor neurons (LMN) from the spinal cord and brain stem and upper motor neurons (UMN) from the motor and premotor cortex. There can also be extra-motor involvement [1]. The pathology of ALS includes aggregation of protein within cells [2]. The protein accumulation spreads to other regions as disease progresses [3]. There is a need for non-invasive biomarkers as a means of monitoring progression of patients with ALS, for prognosis and for use in clinical trials. Possible biomarkers include clinical, neurophysiological, blood and imaging measurements. An imaging biomarker is desirable because it can show changes in brain structure that can be related to the pathology of disease.

Magnetic resonance imaging (MRI) has an advantage of being non-invasive, and able to quantify brain structure and to make inferences about tissue pathology. Cross sectional MRI studies in ALS have shown abnormalities in voxel–based morphometry (VBM) and diffusion weighted imaging (DWI) [4‐7]. Because ALS is a disease that shows degeneration of white matter fiber tracts, particularly the corticospinal tract (CST), there have been studies that use diffusion tensor imaging (DTI). DTI can be analyzed with Tract-Based Spatial Statistics (TBSS) [8] or region of interest (ROI) studies, which can be can be performed by manual or semi-automated methods. Manually drawn ROIs have been considered to have problems with reproducibility, but have the advantage of ease of use.

All DTI metrics have been found to reflect microstructural pathology. What we know about DTI metrics is largely based upon empirical studies that investigate how these measures correlate with the underlying pathology. Quantitative FA measures are believed to reflect changes in the myelination, fiber density and packing [9]. Evidence suggests that fractional anisotropy (FA) is sensitive to the WM integrity but is not very specific and cannot distinguish between different pathological processes [10‐12]. Mean diffusivity (MD) is suggested to be influenced by free water diffusion in cellular space, such as oedema, which is increased in neurodegenerative diseases [12]. Axial diffusivity (AD) is thought to reflect demyelination in white matter tracts. Axial integrity of axons and can be reflected as either increased or decreased AD when WM is destroyed by any other process [13]. It has not yet been well-defined whether radial diffusivity (RD) is specific to demyelination and could also be related to be due to; the occurrence of inflammation, or to axonal damage. RD was found to be high in neurodegenerative disease [14]. RD has been shown to correlate with demyelination and dysmyelination in several pathological models [15, 16].

Cross-sectional DTI studies have consistently shown reduction in FA in the corticospinal tract (CST) in ALS patients compared to controls [17‐20]. In contrast, longitudinal DTI studies of ALS have shown conflicting findings, but most studies show DTI abnormalities early in disease, but little progression over time [7, 21‐26].

We have previously reported a study comparing scans 6 months apart, and found little difference in DTI measures [24]. In this study we have focused on the CST, because degeneration of the CST is a cardinal feature of the pathology in ALS. To examine whether changes can be detected after a longer interval, we now report a 12 month serial study of patients with ALS, with the aim of determining whether there are changes over time and the second aim was to compare semi-automated ROI with manual ROI methods.

This study shows that there is no significant change in DTI along CST in the selected ROIs in ALS patients over 12 months. Manual and automated ROI methods gave comparable results.

The underlying pathology of ALS is death of upper motor neurons, degeneration of the CST, and death of lower motor neurons, with eventual gliosis of the CST in the spinal cord. In the present study, we have examined changes in DTI measures, with the hypothesis that there would be degeneration over time seen as loss of FA in WM tracts. We used TBSS for an overview of all the fiber tracts and ROI of the PLIC and CST to determine whether MRI can show changes in these motor pathways. We find that there is no significant change in DTI over a 12 month period. We also find that the PLIC shows better correlation between manual and automated ROI measures than pons.

Longitudinal DTI studies in ALS are challenging and the findings are variable, but generally concur that fiber tract degeneration has typically taken place by the time the diagnosis is confirmed, with limited additional WM changes in later phase of the disease [39]. An older longitudinal DTI study with only six noncollinear directions have reported positive findings but on a very small sample size [40]. Our own previous study found little change in DTI over 6 months [24]. In the present study we examined changes over a 12 month period, to see whether DTI changes appear over a longer time interval.

Data were investigated at three different time-points, 6 months apart. With TBSS there was some evidence of change over time in motor pathways. The ROI analyses were used to test our hypothesis that there would be changes in the FA in the CST at pons and PLIC. There were no significant differences in the average FA between time-point using aROI. However, in this study some patients showed loss of FA over time while others did not. This difference between TBSS and ROI has been noted previously, in a study that found ROI methods to show less significant results than TBSS methods [41].

There are some reasons why there would be little change in FA over time. Firstly, ALS patients are heterogeneous, with some patients showing prolonged survival [42‐44]. In our cohort, some of the patients had prolonged survival, indicating slowly progressive disease, and therefore would not be expected to show rapid change in fiber tracts. It may be that patients with rapidly progressing disease would show DTI changes.

Next, it is likely that after a fiber tract has completely degenerated, there will be no further change in DTI. Some of our patients were in later stages of the disease, when fiber tracts could be completely degenerated. If loss of fiber tracts leads to decline in FA, then the decline would be expected to cease at this point.

However, it is also possible that, even early in disease, the degeneration in the motor tracts is substantially advanced. Other studies have reported little change in DTI over time [7, 45]. These authors suggested that this indicates that upper motor neuron degeneration occurs early in the disease. Our previous MRI study showed that UMN degeneration occurs early [46], and neurophysiological studies also suggest this [47]. This could mean that changes in FA occur early in disease. It may be that it is better to measure spread of DTI changes to new regions rather than follow changes in the CST [48]. Heterogeneity of clinical features in ALS patients is also an issue and subgroup analysis according to features such as the presence of cognitive impairment would be useful.

In the current study we compared atlas based and manual ROI. The manual methods gave a wider spread of values of FA. We found better agreement from the PLIC than the CST in the pons. Linear regression showed a significant relationship between the manual and atlas based methods for the PLIC but not the pons. This can be explained by the crossing WM pathways in the pons which can lead to a substantial drop of FA in the map. Using Spearman’s correlation, there was a significant correlation of the values for the left PLIC. Manual ROI has been criticized for being operator dependent [49]. However, we found that both methods are still comparable which is in support of previous studies [50, 51]. Rigorous published work have compared different automated methods [52] and test-retest reproducibility [53] in segmentation approaches suggests that atlas-based methods are the best [54].

Laterality was not the aim in the current study. However, it is possible that our lack of significant difference in pons was due to not only the presence of crossing fibers but also to the challenges in delineating this structure, due to its small size, and difficulty discerning its boundaries within the brain stem. However, this variability in pons does limit the interpretation of our findings.

Previous studies have reported that although inter-site DTI measurement variability was found to be quite high [55], the intra-subject variation was less than 1% in different brain regions using DTI [56]. Our patients showed variation of more than 1%, therefore changes we observed are unlikely to be caused by measurement error alone. Compared to previous studies, the ICC from the current study shows that the manual ROI has a good reproducibility and the regions were drawn consistently [57].

Limitations of our study include the small sample size. More investigations are needed to characterize different ALS stages with a larger group of cohorts.

In conclusion, we found little evidence of progression of ALS using DTI. It may be that there is little further change after diagnosis because the pathological changes are complete. This study adds more information about the use of DTI in ALS longitudinal studies.