Alteration of gray matter texture features over the whole brain in medication- overuse headache using a 3-dimentional texture analysis

DOI: https://doi.org/10.1186/s10194- 017- 0820- 4

Abstract-Summary The imaging processing included two steps: gray matter (gray images) segment and a 3-dimensional texture features mapping.

Voxel-based gray-level co-occurrence matrix (VGLCM) was performed to mea- sure the texture parameters mapping including Contrast, Correlation, Energy, Entropy and inverse difference moment (IDM).

The texture parameters of increased Contrast and Entropy, decreased Energy and

IDM were identified in cerebellar vermis of MOH patients compared to NCs.

Increased Correlation located in left dorsolateral periaqueductal gray (L-dlPAG), right parahippocampal gyrus (R-PHG), and left middle frontal gyrus (L-MFG) and decreased Correlation located in right superior parietal lobule(R-SPL).

The altered textures in gray matter related to pain discrimination and modula- tion, affective and cognitive processing were helpful in understanding the pathogen- esis of MOH.

Texture analysis using VGLCM is a sensitive and efficient method to detect sub-

tle gray matter changes in MOH.

Extended: The texture parameters of increased Contrast and Entropy, decreased Energy and IDM in cerebellar vermis of MOH patients indicated that the texture fea- ture of vermis in MOH was more heterogeneous and complicated in MR T1 images. Texture analysis using improved VGLCM method was sensitive and efficient in

detecting subtle structural changes over the whole brain in MOH.

Background Recent studies presented that texture analysis had widely been applied in the clini- cal practice, such as rectal cancer [574, 575], hepatic hemangioma [576], mild cog- nitive impairment [577], glioblastoma [578], and etc In our previous study [579],

3.5 Medication Overuse and Addiction

593

we performed a pilot texture feature analysis for periaqueductal gray (PAG) using a 2-dimentionalgray-level co-occurrence matrix (2D–GLCM) and primary results suggested increased Contrast presented in MOH patients.

In recent documents [580, 581], 3-dimentional(3D) texture analysis technique were reported, and it had some valuable merits, such as voxel-based analysis over the whole brain, and not influenced by the size and the direction.

The optimization of the 3D texture calculation would improve the clinical appli-

cation of this texture technique.

We hypothesize that there’s altered texture feature of gray matter in MOH

patients.

The five texture feature maps were generated and performed with voxel-based analysis over the whole brain to identify the brain regions with abnormal texture changes in MOH.

Methods The key settings of the texture maps calculation were the radius R of a spherical region of interest around each voxel and the maximum distance d for the tex- ture points.

The positive clusters were saved as masks to extract the texture values, and then were performed correlation analysis with the clinical variables (including VAS score, disease duration, HAMA score, HAMD score, MoCA score).

The data with normal distribution was described as mean ± standard deviation, and performed with independent two-sample t test and Pearson correlation analysis. The data without normal distribution was described as median (P25, P75), and

performed with Mann-Whitney U test and Spearman correlation analysis.

Voxel-based texture features analysis over the whole brain was performed between MOH and NC, and two-sample t-test design model was selected to identify the brain regions with significant altered texture features in MOH.

Results Significant increased HAMA and HAMD scores and decreased MoCA score were identified in MOH compared with NC (P < 0.05).

There were no brain regions with decreased Contrast in MOH.

Discussion The texture map analyses in the present study revealed altered texture features in cerebellar vermis, left cerebellum, left dorsolateral periaqueductal gray (L-dlPAG), right parahippocampal gyrus (R-PHG), left middle frontal gyrus (L-MFG) and right superior parietal lobule(R-SPL).

The texture parameters of increased Contrast and Entropy, decreased Energy and IDM in cerebellar vermis of MOH patients indicated that the texture feature of ver- mis in MOH was more heterogeneous and complicated in MR T1 images.

Our study found significantly higher HAMD score in MOH patients and a nega- tive correlation between the Correlation value of R-PHG and HAMD score, which further identified the possible anatomic basis for emotional change and dependence behavior of MOH patients.

594

3 Diagnosis

Only five texture features were calculated in this study, and more texture features may be considered to screen the significant texture features for MOH patients in the future.

Conclusions This study revealed altered texture feature of several brain regions in MOH patients, which may reflect the neural plasticity of pain discrimination and modulation, affec- tive and cognitive processing in MOH and were helpful in understanding the patho- genesis of MOH.

Texture analysis using improved VGLCM method was sensitive and efficient in

detecting subtle structural changes over the whole brain in MOH.

Acknowledgement A machine generated summary based on the work of Chen, Zhiye; Chen, Xiaoyan; Chen, Zhiqiang; Liu, Mengqi; He, Huiguang; Ma, Lin; Yu, Shengyuan. 2017 in The Journal of Headache and Pain.

FRAMES protocol versus simple advice for medication-overuse headache: a prospective, randomized, controlled clinical trial