Hoai-Thu Nguyen^{1}, Thomas Grenier^{2}, Benjamin Leporq^{2}, Loïc Bey^{1,3}, Magalie Viallon ^{1,3}, and Pierre Croisille^{1,3}

This study investigates the local changes in femoral bone marrow in a longitudinal MRI dataset of mountain ultra-marathon runners acquired during the Tor des Géants 2014 based on the MRI quantitative metrics with a rigorous statistical analysis procedure. The results highlight the different characteristics of different quantitative metrics that provides multiple insights into the data and open various perspectives for further analysis.

Our data
included 3D Water images (2-points Dixon, 640 axial slices) MR images collected
from 20 finishers from the Tor Des Géants MUM 2014 (D+24000, 330km), as well as
derived quantitative maps^{5} of T2* and χ for all 3
acquisition time points (Pre - baseline before the race, Post - at the finish
line and Post+3 - 48h after the race).

Based on the automatic
segmentation of femur^{6}, we applied an automatic
segmentation refinement process to extract the exact segmentation of the bone
marrow (Figure 1) on both legs of each runner. In order to monitor the
longitudinal changes in the femoral bone marrow, we used our final segmentation
to extract statistical data from the quantitative maps, and then we performed a
rigorous statistical test procedure to identify significant difference among
the three-time points. The significance level was set at P ≤ .05 for all the
statistical tests with the null hypothesis is that the mean values of the
observed data are the same among time points. Depend on the normality of the
data, we employed either repeated measures ANOVA or Friedman test for global
effect test, dependent t-test or Wilcoxon signed-rank test for pairwise tests.
The P-values obtained for pairwise tests were adjusted with the Bonferroni method
for multiple comparisons.

**Results & Discussion **

The statistical tests (Figure 2) showed a statistically significant time effect on most of the image features of T2* (6/7 features) with significant differences between Pre and Post and between Pre and Post+3, which implied that the time of 2 days was not enough for the bone marrow to return to its initial state. In the meantime, for the χ quantitative maps, only the histogram’s skewness displayed statistically significant longitudinal changes.

Looking
closer to the data (Figure 3), we discovered that the T2* of all the runners
share similar longitudinal patterns with relatively small inter-individual
variation while we observed very different behaviors in the data of χ. Therefore,
we split the set of runners into two categories: one with positive variation during
the race (between Pre and Post) and one with negative variation. For both
categories, the results displayed significant global time effect with a statistically
significant difference between Pre and Post, which indicates that there were at
least two different populations among our runners with different biological
response to the race (Figure 4). The change in χ could be related to the changes
in the concentration of red/white blood cells in the bone marrow. Spiropoulos
et al.^{4} also stated
that the different exercise conditions and lengths could lead to either the
increase or the decrease of the amount of red blood cells. While our runners were
all experienced, they have a wide age range and large morphological difference,
which might explain the variation in behaviors of the bone marrow.

We studied the femoral bone marrow variation of MUM runners with the help of two types of MRI quantitative maps (T2* and χ). T2* highlighted statistically significant time effect for most of the statistical tests meaning an ineligible change in the water concentration in the bone marrow of runners. Furthermore, χ maps revealed two populations with different behaviors in the bone marrow after the race. This suggests that the T2* metric can provide a global view to the dataset while the χ metric is more sensitive to the variation among the subjects.

Further analysis is in process while taking in to account biological makers, more advanced feature extracted from MR images and intra-individual variations among runners.

1. Millet GP, Millet GY. Ultramarathon is an outstanding model for the study of adaptive responses to extreme load and stress. BMC Med. 2012;10(1):1-3. doi:10.1186/1741-7015-10-77.

2. Gabay C, Kushner I. Acute-Phase Proteins and Other Systemic Responses to Inflammation. N Engl J Med. 1999. doi:10.1056/NEJM199902113400607.

3. Wright TM. Cytokines in acute and chronic inflammation. Front Biosci. 1997. doi:10.2741/A171.

4. Spiropoulos A, Goussetis E, Margeli A, et al. Effect of inflammation induced by prolonged exercise on circulating erythroid progenitors and markers of erythropoiesis. Clin Chem Lab Med. 2010. doi:10.1515/CCLM.2010.034.

5. Leporq B, Le Troter A, Le Fur Y, et al. Combined quantification of fatty infiltration, T 1-relaxation times and T 2*-relaxation times in normal-appearing skeletal muscle of controls and dystrophic patients. Magn Reson Mater Physics, Biol Med. 2017;30(4):407-415. doi:10.1007/s10334-017-0616-1.

6. Gilles B, De Bourguignon C, Croisille P, et al. Automatic segmentation for volume quantification of quadriceps muscle head: a longitudinal study in athletes enrolled in extreme mountain ultra-marathon. In: ISMRM: International Society for Magnetic Resonance in Medicine. ; 2016.

Figure
1:
Example of T2* and χ maps of the right leg of the same subject with the
segmentation of the bone marrow at the three MR acquisition time points of the
race.

Figure
2:
P-values of statistical tests in our longitudinal analysis of the entire data
set. A P-value inferior to .05 indicates a significant change between two time
points. Abbreviations: BM – Bone Marrow, r – right, l – left.

Figure
3:
Variation of the mean of quantitative metrics in the femoral bone marrow during
the three studied time points. A distinct color represents each runner; the
gray bars represent the average values.

Figure
4:
P-values of statistical tests in our longitudinal analysis on the χ map. On the
left, we considered only the individuals with a positive variation of the mean
of χ between Pre and Post. On the right, we considered the individuals with
negative variation. A P-value inferior to .05 indicates a significant change
between two time points. Abbreviations: BM – Bone Marrow, r – right, l – left.