Correlation Study Between Spinal Cord Function, Spinal Cord Morphology and Cervical Spine Alignments in Patients With Hirayama Disease

Abstract

Objective

This study aimed to explore the relationship between diffusion tensor imaging (DTI) parameters, cervical spine alignments, and spinal cord morphological parameters in patients with Hirayama disease (HD).

Methods

In this retrospective cohort study, 41 HD patients were recruited from the Huashan hospital from July 2017 to November 2021. Patients received X-rays, conventional magnetic resonance (MR), and DTI scans in flexion and neutral positions. The DTI parameters assessed were calculated using the region of interest (ROI) method. Paired t-tests were performed on the DTI parameters of neck flexion and neutral position. Cervical spine alignments, including flexion and neutral Cobb angles, were measured, and range of motion (ROM) was calculated. Spinal cord morphological parameters were measured, including spinal cord atrophy (SCA) and loss of attachment (LOA). Spearman’s correlation analysis between DTI parameters, cervical spine alignments, and spinal cord morphological parameters was performed.

Results

In comparing DTI parameters, segments of the C3/4, C4/5, C6/7, and lower cervical spine were significantly different, while segments of C5/6 were not significantly different. In Spearman’s correlation analysis, the flexion Cobb angle was significantly correlated with the fractional anisotropy (FA) value (R² = .111, P = .033) and apparent diffusion coefficient (ADC) value (R² = .119, P = .027). Flexion FA values were correlated with SCA in C4/5 (R² = .211, P = .003), C5/6 (R² = .454, P < .001), and C6/7 (R² = .383, P < .001) while flexion ADC values were correlated with SCA in the C4/5 (R² = .178, P = .006), C5/6 (R² = .388, P < .001) and C6/7 segments (R² = .187, P = .005).

Conclusion

The DTI parameters were correlated with the flexion Cobb angle and the SCA. These data support the dynamic cervical flexion compression hypothesis and indicate that the degree of SCA may be used to assess the condition of HD patients quantitatively.

Introduction

Hirayama disease (HD), also known as juvenile distal muscular atrophy, was first reported by Japanese physician Hirayama Keizo in 1959. ¹ Previous research has indicated that HD is prevalent in Asian countries such as Japan, China, and India due to its higher incidence in Asian adolescents.^2-4 HD was previously thought to be a benign and self-limiting disease. ⁵ Consequently, many spine surgeons have historically disregarded it due to its rarity. However, accumulating evidence shows that HD patients exhibit persistently progressive symptoms and ineffective responses to conservative treatment, and the incidence and prevalence of the disease are on the rise.^6-10 Although research on HD remains limited to case reports in Europe and the United States, the number of published studies researching this disease is increasing.^10-14 Therefore, spine surgeons must improve their understanding of the disease.

Although the pathogenesis of HD is not fully understood, various hypotheses have been proposed, with the dynamic cervical flexion compression hypothesis being the most significant. ⁵ According to this hypothesis, repeated or continuous flexion of the cervical spine leads to the movement of the posterior dural sac forward to the subjacent lamina, resulting in spinal cord compression, microcirculation disturbances, and chronic injury in the anterior horn of the spinal cord. ¹⁵ Based on this hypothesis, many physicians began to study HD patients using various imaging methods, such as dynamic cervical spine X-ray and conventional magnetic resonance (MR), to explore the cervical spine alignment,^16,17 and spinal cord morphology, respectively. ¹⁸

Diffusion tensor imaging (DTI) can non-invasively detect the structure of neurons and indirectly reflect neurologic function. Previously, our group applied DTI to study HD, completing DTI scans of 17 HD patients. The results revealed significant differences in the DTI parameters of the patients as compared to those of healthy volunteers, and significant differences were also observed in the DTI parameters of the patients in both the neck flexion and neutral positions. ¹⁹ Despite these findings, no studies have examined the correlation between cervical spine alignments, spinal cord morphology, and spinal cord function. This study aims to investigate this relationship by utilizing dynamic X-ray, conventional MR, and spinal cord DTI.

Materials and Methods

Patients

Patients were recruited from July 2017 to November 2021 and were required to meet clear diagnostic criteria for HD, as defined by clinician-led international guidelines for diagnosing and treating Hirayama disease established in 2020. ²⁰

The study’s inclusion criteria consisted of (a) the presence of symptoms such as unilateral intrinsic hand muscle atrophy, cold paralysis, and extensor tremors; (b) the observation of spinal cord atrophy (SCA) and loss of attachment (LOA) on the sagittal plane on MR T2-weighted imaging (T2WI) when the neck was flexed; and c) the presence of segmental localized nerve damage in the anterior horn or anterior root of the spinal cord as shown by electromyography (EMG). Patients with traumatic brain injury, epilepsy, Parkinson’s disease, or other neurological diseases, and if they experienced claustrophobia syndrome, had internal metal instrument placement or had other contraindications to MR examinations, were excluded. Ultimately, 41 patients met the inclusion criteria and were enrolled in the study.

We also classified our HD patients according to the Huashan clinical classification system. ²⁰ This system was able to facilitate daily routine medical practice on Hirayama disease as well as oral and written communications during academic conferences. It has also been proven to be reproducible and reliable. ²¹ According to the system, HD patients can be divided into type I, type II and type III. Patients of type I have typical symptoms such as unilateral muscular atrophy without pyramidal tract signs or sensory disturbances. Patients of type II have typical symptoms accompanied by sensory disturbances or pyramidal tract signs. Patients of type III is atypical HD involving the proximal muscles of the upper limbs or bilateral symptoms. Finally, our patients were divided into: 24 cases of type I, 13 cases of type II, and 4 cases of type III.

Radiographic Assessment

To collect cervical spine alignments, neutral Cobb angle and flexion Cobb angle were measured. In the patients’ lateral radiographs, tangents were made to the lower endplates of C2 and C7, and perpendiculars were made to the tangents. The acute angle formed by the intersection of the two perpendiculars was considered the Cobb angle (as shown in Figures 1A and 1B). If the two tangents intersected behind the spine, the Cobb angle was positive, whereas if it intersected in front of the spine, it was negative. The range of motion (ROM) was calculated as the difference between the neutral and flexion Cobb angles. Two physicians measured all parameters independently, and the final result was an average of the two measurements.

Figure 1.

Measurement of cervical spine alignments and spinal cord morphological parameters. Cobb angle of C2/7 were measured in (A) neutral position and (B) flexion position based on Cobb's method. The spinal cord morphological parameters were measured according to the pattern map (C). First, we need localizing in the sagittal plane (D). Next, we can measure the degree of SCA (E) and the degree of LOA (F) DTI, diffusion tensor imaging; SCA, spinal cord atrophy; LOA, loss of attachment.

Conventional MR Assessment

Conventional MR images were obtained using a 3.0 T MRI scanner (MAGNETOM Verio, Siemens, Germany) in the neck flexion position. T2WI were acquired using a fast spin echo (FSE) sequence with the following parameters: repetition time (TR)/echo time (TE) of 2440/122.7 ms, layer thickness/layer spacing of 3/1 mm, the sagittal field of view (FOV) of 300 × 300 mm², cross-sectional FOV of 240 × 240 mm², and matrix of 128 × 128.

The collected spinal cord morphological parameters included the degree of SCA and LOA. The measurement was limited to the lower cervical spine since the morphological changes in HD patients are mainly located in the lower cervical spine (below C4/5, including C4/5). The anteroposterior diameter of the spinal cord was recorded as x, and the left to right diameter of the spinal cord was recorded as y. The degree of SCA was calculated as x/y (Figure 1C and E). Similarly, the distance from the spinal cord’s posterior end to the spinal canal’s posterior wall was recorded as a, and the distance from the spinal cord’s anterior end to the spinal canal’s posterior wall was recorded as b. The degree of LOA was calculated as a/b (Figure 1C and F). Each parameter was measured on two cross-sectional images, and the average was calculated. Two physicians measured all parameters independently, and the final measurement result was the average of the two physicians.

DTI Assessment

DTI images were acquired using a 3.0 T MRI scanner (MAGNETOM Verio, Siemens, Germany). Each patient was scanned twice, once in the neck flexion position and once in the neutral position. Anatomical images were obtained through sampling perfection with the application-optimized contrasts using different flip angle evolutions (SPACE) sequences. Diffusion images were obtained using a single-shot echo planar imaging (SSEPI) sequence. The imaging parameters were set as follows: b-value of 0/500 s/mm², diffusion gradient direction of 30, repetition time (TR)/echo time (TE) of 2900/61 ms, slice thickness/slice spacing of 3/1 mm, sagittal field of view (FOV) of 300 × 300 mm², cross-sectional FOV of 240 × 240 mm², matrix of 128 × 128, and voxel size of 2.0 × 2.0 × 2.0 mm³.

The DTI images were post-processed and reconstructed using Siemens Healthcare AG in “neuro 3D” mode. The DTI parameters, including fractional anisotropy (FA) and apparent diffusion coefficient (ADC) values, were obtained using the region of interest (ROI) method. The fusion image of the anatomical and diffusion images was used to locate the mid-sagittal plane, and a circular ROI was drawn in the spinal cord area, ensuring that each ROI did not exceed the spinal cord and contained at least 2 voxels. Measurements were taken in the neck neutral position of each patient (Figures 2A and B), and the same method was used to measure each patient’s neck flexion position (Figures 2C and D). Each measurement was also performed on the adjacent sagittal planes on the left and right sides of the mid-sagittal plane, and the average value was taken as the parameter. Using the above method, the DTI parameters were obtained for each segment of C2/3-C6/7, and the worst two segments of the lower cervical spine (C4/5-C6/7) were calculated to represent the lower cervical spine parameters. All measurements were completed independently by two physicians, and the final result was the average of the two physicians.

Figure 2.

Measurement of DTI parameters in neutral (A & B) and flexion (C & D) position based on ROI method. In the Siemens post-processing workstation, the anatomical images and the diffusion images are fused (B & D). A circular area is drawn at the position of the spinal cord. The computer will automatically generate the FA value and ADC value of this area (A and C) DTI, diffusion tensor imaging; ROI, region of interest; FA, fractional anisotropy; ADC, apparent diffusion coefficient.

Statistical Analysis

All statistical analyses were performed using IBM SPSS 23.0 (IBM Corp, Armonk, NY, USA). Demographic data were analyzed by one-way ANOVA. Sex was tested by chi-square test. The flexion and neutral DTI parameters of each segment and the lower cervical spine were tested by paired-sample t test. Spearman’s correlation analysis was used to determine the correlations between flexion DTI parameters and cervical spine alignments and between flexion DTI parameters and spinal cord morphological parameters.

Results

The classification and demographics of the patients were summarized in Table 1. We found that the duration of disease (P = .007) and the age (P = .038) were statistically significantly different among the three groups of patients, which is consistent with previous reports in the literature. ¹⁷ Figures 3 and 4 respectively showed the basic data of patients with type I and type II HD.

Table 1.

Comparison of demographic data of patients with different types of Hirayama disease.

	Type Ⅰ (N = 24)	Type Ⅱ (N = 13)	Type Ⅲ (N = 4)	P-Value
Height (cm)	171 ± 6	173 ± 6	173 ± 5	.544
Weight (kg)	63 ± 11	63 ± 9	68 ± 6	.607
Sex (male/total)	23/24	12/13	4/4	.646
Age (year)	18 ± 3	19 ± 3	24 ± 6	.038*
Age of onset (year)	16 ± 4	17 ± 2	19 ± 5	.225
Duration of disease (month)	21.7 ± 18.8	28.0 ± 19.5	56.0 ± 15.8	.007**

Values are presented as mean ± standard deviation.

*P < .05, **P < .01, ***P < .001.

Figure 3.

A 16-year-old adolescent male with type I HD presented with progressive left distal upper extremity muscle weakness and atrophy for more than 6 months (A). EMG showed chronic neurogenic impairment involving the C7-T1 myotome. X-rays showed cervical alignments imbalance (B) and significant increase in flexion ROM (C). T2WI in the flexion position showed SCA and LOA (D & E). The patient received spinal DTI scan (F & G) and ACDF surgery (H & I). HD, Hirayama disease; EMG, electromyography; T2WI, T2-weighted imaging; SCA, spinal cord atrophy; LOA, loss of attachment; DTI, diffusion tensor imaging; ACDF, anterior cervical discectomy and fusion.

Figure 4.

A 22-year-old adolescent male with type Ⅱ HD presented with progressive right distal upper extremity muscle weakness and atrophy for more than 6 months (A). EMG showed chronic neurogenic impairment involving the C7-T1 myotome. X-rays showed cervical alignments imbalance (B) and significant increase in flexion ROM (C). T2WI in the flexion position showed SCA and LOA (D & E). The patient received spinal DTI scan (F & G) and ACDF surgery (H & I). HD, Hirayama disease; EMG, electromyography; T2WI, T2-weighted imaging; SCA, spinal cord atrophy; LOA, loss of attachment; DTI, diffusion tensor imaging; ACDF, anterior cervical discectomy and fusion.

Two male adolescents, aged 16 and 22, presented with progressive, unilateral distal upper extremity muscle weakness and atrophy that had persisted for over six months. EMG data indicated chronic neurogenic impairment involving the C7-T1 myotome. X-ray data indicated a significant increase in flexion ROM and cervical alignment imbalance in the neutral position. MR T2WI in the flexion position showed SCA and LOA between the posterior dural sac and the subjacent lamina. One of the patients even presented with the snake eye sign (Figure 4E). They received a spinal DTI scan and anterior cervical discectomy and fusion (ACDF) surgery considering their progressive worsening symptoms. Thirty-six months follow-up showed that patients’ symptoms had stopped progressing.

Comparison of DTI Parameters Between cervical flexion and Neutral Positions

First, the flexion and neutral parameters were compared (Table 2). Statistical analysis revealed that the FA values for the C2/3 (flexion vs neutral: .513 ± .094 vs .584 ± .107, P = .002), C3/4 (flexion vs neutral: .502 ± .097 vs .605 ± .125, P < .001), C4/5 (flexion vs neutral: .439 ± .095 vs .557 ± .135, P < .001), C6/7 (flexion vs neutral: .399 ± .109 vs .456 ± .105, P = .027) segments and lower cervical spine (flexion vs neutral: .387 ± .092 vs .432 ± .098, P = .026) were significantly different. Similarly, the comparison of ADC values showed that the C3/4 (flexion vs neutral: 1.542 ± .337 vs 1.305 ± .344, P < .001), C4/5 (flexion vs neutral: 1.825 ± .466 vs 1.489 ± .440, P < .001), C6/7 (flexion vs neutral: 2.161 ± .420 vs 1.866 ± .485, P = .005) segment and the lower cervical spine (flexion vs neutral: 2.217 ± .472 vs 1.991 ± .423, P = .012) were significantly different. In the C5/6 segment, neither the FA nor the ADC values significantly differed. However, by analyzing the data, we found that the C5/6 segment has a low FA value and a high ADC value, even in the neutral position. Therefore, these data suggest severe damage to the spinal cord in this segment, even in the neutral position.

Table 2.

Comparison of DTI parameters between cervical flexion and neutral position in HD patients.

Segment	Flexion FA value	Neutral FA value	P-Value	Flexion ADC value	Neutral ADC value	P-Value
C2/3	.513 ± .094	.584 ± .107	.002**	1.432 ± .263	1.327 ± .323	.109
C3/4	.502 ± .097	.605 ± .125	<.001***	1.542 ± .337	1.305 ± .344	<.001***
C4/5	.439 ± .095	.557 ± .135	<.001***	1.825 ± .466	1.489 ± .440	<.001***
C5/6	.419 ± .111	.438 ± .130	.395	2.149 ± .640	1.986 ± .534	.114
C6/7	.399 ± .109	.456 ± .105	.027*	2.161 ± .420	1.866 ± .485	.005**
Lower cervical spine	.387 ± .092	.432 ± .098	.026*	2.217 ± .472	1.991 ± .423	.012*

Values are presented as mean ± standard deviation.

DTI, diffusion tensor imaging; HD, Hirayama disease; FA, fractional anisotropy; ADC, apparent diffusion coefficient.

We used the mean of the two segments with the smallest FA value among the three segments C4/5-C6/7 to represent the FA value of the lower cervical spine. In the same way, we get the ADC value of lower cervical spine by selecting the two largest values.

*P < .05, **P < .01, ***P < .001.

Correlation Between DTI Parameters and Cervical Spine Alignments

Next, the patient’s flexion lower cervical spine DTI parameters and cervical spine alignments were assessed (Table 3). A spearman’s correlation analysis revealed that the flexion Cobb angle was statistically significantly correlated with the FA value (R² = .111, P = .033) and ADC value (R² = .119, P = .027), while the neutral Cobb angle and ROM did not correlate with the DTI parameters (Figure 5).

Table 3.

Flexion lower cervical spine DTI parameters and cervical spine alignments in HD patients.

Flexion FA value	Flexion ADC value	Neutral Cobb angle (°)	Flexion Cobb angle (°)	ROM (°)
.387 ± .092	2.217 ± .472	5.87 ± 9.63	-33.02 ± 8.80	38.89 ± 8.74

Values are presented as mean ± standard deviation.

DTI, diffusion tensor imaging; HD, Hirayama disease; FA, fractional anisotropy; ADC, apparent diffusion coefficient; ROM, range of motion.

Figure 5.

Spearman’s correlation analysis between cervical spine alignments and DTI parameters. The flexion cobb angle is significantly correlated with flexion FA value (R2=0.111, P=0.033) and flexion ADC value (R2=0.119, P=0.027). However, other cervical spine alignments are not significant correlated with DTI parameters DTI, diffusion tensor imaging; FA, fractional anisotropy; ADC, apparent diffusion coefficient.

Correlation Between DTI Parameters and spinal Cord Morphological Parameters

Next, the patient’s flexion DTI and spinal cord morphological parameters of each segment of C4/5-C6/7 were assessed (Table 4). Spearman’s correlation analysis revealed significant correlations between the flexion FA value and the SCA in the C4/5 (R² = .211, P = .003), C5/6 (R² = .454, P < .001) and C6/7 (R² = .383, P < .001) segments (Figure 6A-C). Additionally, the flexion ADC values were also correlated with SCA in the C4/5 (R² = .178, P = .006), C5/6 (R² = .388, P < .001), and C6/7 (R² = .187, P = .005) segments (Figures 6D-F). However, FA and ADC were not correlated with LOA in any segment (Figure 7).

Table 4.

Flexion DTI parameters and flexion morphological parameters of each segment of lower cervical spine in HD patients.

Segment	Flexion FA value	Flexion ADC value	SCA	LOA
C4/5	.439 ± .095	1.825 ± .466	.434 ± .074	.531 ± .086
C5/6	.419 ± .111	2.149 ± .640	.381 ± .064	.604 ± .066
C6/7	.399 ± .109	2.161 ± .420	.388 ± .077	.621 ± .066

Values are presented as mean ± standard deviation.

DTI, diffusion tensor imaging; HD, Hirayama disease; FA, fractional anisotropy; ADC, apparent diffusion coefficient; SCA, spinal cord atrophy; LOA, loss of attachment.

Figure 6.

Spearman’s correlation analysis between SCA and DTI parameters. SCA is significantly correlated with flexion FA values in C4/5 (R2=0.211, P=0.003), C5/6 (R2=0.454, P=0.003), C6/7 (R2=0.383, P=0.001). SCA is also significantly correlated with flexion ADC values in C4/5 (R2=0.178, P=0.006), C5/6 (R2=0.388, P=0.001), C6/7 (R2=0.187, P=0.005) SCA, spinal cord atrophy; FA, fractional anisotropy;ADC, apparent diffusion coefficient.

Figure 7.

Spearman’s correlation analysis between LOA and DTI parameters. LOA is not significantly correlated with neither flexion FA values nor flexion ADC values. LOA, loss of attachment; FA, fractional anisotropy; ADC, apparent diffusion coefficient.

Discussion

The dynamic cervical flexion compression hypothesis proposes that HD patients have excessive neck flexion activities, which may lead to anterior displacement of the cervical dural sac upon sustained or repeated flexion. This, in turn, facilitates segmental compression of the cervical spinal cord, leading to increased intramedullary pressure and microcirculation disturbances in the anterior horn of the spinal cord, thereby causing local lesions in the spinal cord. ⁵ Based on this hypothesis, numerous studies have explored cervical spine alignments in HD patients. For instance, Xu et al used dynamic X-rays to examine 31 HD patients and 40 healthy volunteers. The authors found that neck flexion ROM significantly differed between the patient and control group. ²² This observation was also verified by Song et al. In that study, Song et al compared cervical spine alignments between 23 HD patients and 21 healthy volunteers. Furthermore, the patients’ parameters before and after surgery were also compared. The authors found statistically significant differences in the alignments between patients and healthy volunteers and between patients before and after surgery. Therefore, they concluded that patients with HD have a cervical spine alignments imbalance, which can be corrected by cervical surgery.^16,17 Additionally, Lu et al compared the biomechanical characteristics of the cervical spine in HD patients and healthy volunteers using finite element models and found that HD patients had greater cervical spine stress, increasing their susceptibility to cervical degenerative disease. ²³ Despite these findings, it is crucial to note that the primary lesion in HD patients is located in the spinal cord, and the existing studies have not yet established an association between the cervical spine alignments imbalance and impaired spinal cord function.

Previous studies have shown that DTI parameters correlate with clinical symptoms in HD and cervical spondylotic myelopathy (CSM).^19,24,25 Our study also found significant differences in DTI parameters between HD patients in neck flexion and neutral position. Furthermore, these results agree with our previous study, ¹⁹ indicating that the spinal cord function of HD patients differs in their flexion and neutral positions. However, compared to our previous work, the sample size of this study was larger, and the statistical difference was more pronounced. Despite the measured parameters in the C5/6 segment being similar, the neutral parameters of C5/6 were worse than the flexion parameters of many segments, which may indicate that the C5/6 spinal cord function of the sample was very severely damaged. Additionally, these data indicated that even the C5/6 neurons were irreversibly damaged. In the correlation analysis between DTI parameters and cervical spine alignments, we found that both FA value (R² = .111, P = .033) and ADC value (R² = .119, P = .027) were statistically significantly correlated with the flexion Cobb angle, which indicated that cervical spine alignments imbalance of HD patients might lead to spinal cord dysfunction.

Patients with HD usually show anterior displacement of the dural sac, asymmetric SCA, and LOA in the sagittal image of the neck flexion T2WI.^3,26-28 However, HD patients differ in their tolerance and responsiveness to spinal cord compression. ¹⁸ Additionally, no studies have explored the relationship between spinal cord morphological changes and clinical symptoms in HD patients. Therefore, conventional MR imaging only serves a qualitative diagnostic role in HD. Controversy exists about whether conventional MR imaging can quantitatively assess the severity of HD.^15,20,29

DTI can quantitatively assess the neurological function of HD patients and then reflect the severity of the disease. Firstly, DTI parameters have been shown to correlate with clinical symptoms in CSM patients. ²⁴ Secondly, our previous study showed that DTI parameters in HD patients were also correlated with clinical symptoms. ¹⁹ Finally, in this study, we found that the DTI parameters of HD patients in their flexion and neutral positions are statistically significantly different. A statistically significant correlation was also noted between the DTI parameters and the flexion Cobb angle. Based on the above results, a spearman’s correlation analysis was run between the DTI and spinal cord morphological parameters. The analysis found that DTI parameters significantly correlate with the SCA degree. Together, these data indicate that the morphological changes of the spinal cord in HD patients are related to the degree of spinal cord functional damage; that is, the functional damage is manifested by the morphological changes of SCA, which may preliminarily indicate that the degree of spinal cord atrophy has a particular, quantitative evaluation effect. However, our study did not find a correlation between DTI parameters and LOA which may indicate that while LOA played an important role in HD diagnosis, it cannot quantitatively assess HD patients’ condition.

The current study has some limitations that need to be acknowledged. First, due to the elongated shape of the spinal cord, the DTI parameters are obtained by the hand-drawn ROI method. This method means that the surgeon cannot locate the lesion in the grey or white matter or a specific fiber bundle, which may lead to errors because of the mean volume. Second, although DTI can directly detect neuronal microstructural features, it cannot distinguish pathological changes such as edema, demyelination, and Wallerian degeneration. Therefore, DTI parameters are not equivalent to neurological function. Nevertheless, it remains undisputed that DTI parameters are the most effective and sensitive indicators for reflecting spinal cord function.

Conclusion

In the current study, we show that the DTI parameters were correlated with the flexion Cobb angle, and the DTI parameters differed between the flexion and neutral positions of HD patients. These data support the dynamic cervical flexion compression hypothesis, indicating that a smaller Cobb angle in neck flexion indicates worse spinal cord function. Additionally, DTI parameters are correlated with SCA, indicating that the degree of SCA may quantitatively assess the condition of HD patients.

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ORCID iD

Hongli Wang https://orcid.org/0000-0001-5578-0823