A model for estimating the brainstem volume in normal healthy individuals and its application to diffuse axonal injury patients

概要

www.nature.com/scientificreports

*

Traumatic brain injury (TBI) is a major cause of disability, especially among young people, and is a critical
health and socioeconomic problem worldwide1. Some patients with TBI have gross focal injury (FI), which is
characterized by a large focal lesion caused by an external force, whereas others have no visible FI. The latter
TBI cases likely exhibit diffuse pathology, and diffuse axonal injury (DAI) is a major clinical entity of this type2.
DAI is considered a primary cause of consciousness disorder in acute-phase TBI3.
DAI leads to post-acute-phase widespread brain atrophy4–7, which results from Wallerian degeneration and
delayed neuronal cell death, following the process of primary and secondary axotomy because of axonal injury7.
Histopathological observations have revealed that axonal injury in DAI is not evenly diffuse throughout the
brain; rather, it is concentrated in specific brain regions, one of which is the brainstem2,8–11. Given that axonal
injury in the brainstem is a primary factor in the development of immediate consciousness disorder in DAI or
TBI11–13, the post-acute-phase volume of the brainstem in DAI is of clinical interest because it likely reflects
the severity of acute-phase consciousness disorder. However, this issue has not been extensively investigated.
Post-TBI brainstem volume reduction has been reported previously in studies by Kim et al.14 and Cole et al.15;
however, they used tensor-based morphometry or whole-brain voxel-based morphometry analyses in patients
with mild or moderate-severe TBI. Therefore, there is currently limited research on patients across all levels of

Department  of  Psychiatry  and  Behavioral 

Department 
 *

 |           (2023) 13:33  

| 

www.nature.com/scientificreports/

Characteristics

Group A (n = 182)

Group B (n = 47)

P-value

Agea (years)

33 (18–64)

34 (20–68)

0.560b

Sex (male/female)

99/83

24/23

0.683c

Table 1. Demographic characteristics of the participants in Experiment 1. a Median (range). b Groups were
compared with the Mann–Whitney U test. c Groups were compared with the chi-square test.

TBI severity that focuses on the brainstem or its subfields as regions of interest (ROIs). Warner et al.16 reported
volume reductions in the brainstem of patients with traumatic axonal injury, although the exclusion criteria for
FI were relatively liberal (i.e., any focal, mixed, or high-density lesion > 10 ml). Therefore, brainstem volume
changes that are attributed solely to axonal injury have not been fully examined. Brezova et al.17 and Sidaros
et al.5 also reported brainstem volume reduction post-injury; however, the time since injury in the investigated
subjects was < 600 days, when atrophy caused by TBI may still have been in progress18. Taken together, there is
insufficient information on the volume of the brainstem (either the whole brainstem or its subfields) long after
injury or its association with initial injury severity. As such, there are limited normative data available on the
volume of the whole brainstem or any of its subfields. The possibility that the conventional intracranial volume
(ICV) divisor correction or proportions method19,20 is not strictly suitable for brainstem volumetric measures
cannot be ignored. Thus, a new method for evaluating the brainstem volume of any individual in relation to the
healthy volume would be desirable for analysis purposes.
In the present study, we investigated the brainstem volumes of healthy individuals using the following
approach. In Experiment 1, we measured the brainstem volumes of a large number of healthy adults and established a model to estimate the “expected normal healthy volume” (ENHV) of the brainstem in any individual
based on the ICV, age, and sex. In Experiment 2, the first attempt to apply the model to a non-healthy sample was
made; we estimated the percentage difference between the actual brainstem volume and ENHV, and investigated
the association of the ENHV-adjusted brainstem volume with acute-phase injury severity in DAI subjects (who
were mainly in the long-term chronic phase and had no gross FI). The acute-phase severity of TBI is generally
taken to indicate the extent of consciousness disorder, which can be measured by clinical scales such as the Glasgow Coma Scale (GCS) and Japan Coma Scale (JCS), or as the duration of posttraumatic amnesia (PTA). PTA
is a state of confusion that occurs immediately after TBI, in which the patient is disoriented and unable to recall
new memories21–23. In this study, the duration of PTA, which was obtained for every DAI subject, was employed
in the analysis as an indicator of injury severity. We hypothesized that a chronic volume reduction would be
seen in the brainstem, and that this reduction would be associated with the severity of the acute injury. It should
be noted that, considering the global atrophy of the brain that occurs in DAI, we used the whole cerebrum as a
control ROI in all analyses of brainstem volume.
Participants. In total, 185 and 47 healthy subjects participated in Experiment 1, and three
of the 185 healthy subjects were excluded from the analysis because of image processing errors. Finally, the data
of 182 healthy subjects (Group A) and 47 healthy subjects (Group B) were included in the analysis. There was
no significant difference between Group A and B in age or sex. Table 1 shows the demographic characteristics
of the participants.
Creation of the model for estimating ENHV. Figure 1 shows the relationship between each regional brain volume
and age or ICV in the healthy subjects of Group A. There was no significant correlation between age and regional
volume of the whole brainstem (r = 0.093, P = 0.213), medulla (r = 0.029, P = 0.698), pons (r = 0.145, P = 0.050),
or midbrain (r = 0.00046, P = 0.995). However, there was a significant negative correlation between age and the
volume of the cerebrum (r = − 0.332, P = 4.77E-06; Fig. 1a). The volume of each ROI showed a significant positive
correlation with ICV (r = 0.496, 0.567, 0.677, 0.614, and 0.828 for the medulla, pons, midbrain, whole brainstem,
and cerebrum, respectively; P < 1.0E − 11; Fig. 1b). Two-way ANOVA revealed significant interactions between
sex and ICV for the volume of the whole brainstem (F[1, 178] = 8.46, P = 0.004), midbrain (F[1, 178] = 6.49,
P = 0.012), pons (F[1, 178] = 8.60, P = 0.004), and cerebrum (F[1, 178] = 7.11, P = 0.008) but not the medulla (F[1,
178] = 3.44, P = 0.065). Single/multiple regression analyses were performed for each sex group for all ROIs except
the medulla, which was analyzed with the two sex groups combined. The explanatory variables were ICV and
age for the cerebrum and only ICV for the other four ROIs (Table 2). The results of the regression analyses are
shown in Supplementary Table 1, and according to the obtained regression equations, the model equations were
fixed as Eqs. (1), (2), and (3) in Table 2, with the corresponding coefficients and the dummy variable representing
sex shown in Table 3.
Model validation. The intraclass correlation coefficients (ICCs) obtained from the reliability test of the ROIs
are shown in Table 4. For all ROIs, the ICC was > 0.4. Thus, our model had sufficient validity for application of
all five ROIs to Japanese and East Asian adults.
Participants. Twenty-two DAI patients and 60 healthy controls (HCs) participated in Experiment 2. There was no significant difference between the DAI patients and HCs in age or sex. The demo-

 |           (2023) 13:33  | 
Vol:.(1234567890)

www.nature.com/scientificreports/

Figure 1. Relationship between each ROI volume and age (a) or ICV (b) in 182 healthy subjects. (a) There
were no significant correlations between regional volume and age for the whole brainstem (Pearson’s r = 0.093,
P = 0.213), medulla (r = 0.029, P = 0.698), pons (r = 0.145, P = 0.050), or midbrain (r = 0.00046, P = 0.995),
whereas a significant negative correlation was found for the cerebrum (r = − 0.332, P = 4.77E-06). (b) There were
significant positive correlations between regional volume and ICV (r = 0.496, 0.567, 0.677, 0.614, and 0.828 for
the medulla, pons, midbrain, whole brainstem, and cerebrum, respectively; P < 1.0E-11). However, none of the
ROI volumes showed direct proportionality to ICV, reflected by the positive y-intercepts of the regression lines
on the scatterplot with ICV on the x-axis and each regional volume on the y-axis. *Significant at P < 0.05. ROI
Region of interest, ICV Intracranial volume.

ROI

Sample for regression analysis

Explanatory variables

Forms of the model equations

Medulla

Whole sample (two sexes
together) of Group A

ICV

ENHV = a × ICV + b

(1)

Each sex subgroup of Group A

ICV

ENHV = (1 − sex) × (af × ICV + bf ) + sex × (am × ICV + bm )

(2)

Each sex subgroup of Group A

ICV and age

ENHV = (1 − sex) × (af × ICV + bf × age + cf ) + sex × (am × ICV + bm × age + cm ) (3)

Whole brainstem
Pons
Midbrain
Cerebrum

Table 2. Sample and explanatory variables of regression analysis and the model equation form. Each ROI
and its corresponding model equation for estimating healthy volume are shown. The equations had three
forms for the five ROIs. “af × ICV + bf ” or “af × ICV + bf × age + cf ” is the regression equation obtained
in the regression analysis for the female subgroup, and “am × ICV + bm” or “am × ICV + bm × age + cm”
is that for the male subgroup. sex = 1 for males/0 for females in Eqs. (2) and (3). ROI Region of interest, ICV
Intracranial volume, ENHV Expected normal healthy volume. ...

この論文で使われている画像