Evaluation of the corpus callosum using magnetic resonance imaging histogram analysis in autism spectrum disorder

Abstract

Background

Histogram analysis is a texture analysis method that can be used in medical images. Quantitative values of the intensity of images can be obtained with histogram analysis. It aimed to evaluate corpus callosum in magnetic resonance images (MRIs) using histogram analysis of pediatric patients with autism spectrum disorder (ASD) to compare them with healthy controls.

Methods

This study included 29 children with ASD and 29 healthy children with normal brain MRI. High-resolution three-dimensional turbo field echo images were obtained with a 1.5 T scanner device for brain magnetic resonance imaging. On the corpus callosum in the sagittal T1-weighted images obtained, mean gray level density (mean), the standard deviation, median, minimum, maximum, entropy, variance, skewness, kurtosis, uniformity, size % L, size % M, size % U, and percentile parameters were measured.

Results

In ASD patients, mean, standard deviation, maximum, median, variance, entropy, 25%, 75%, 90%, 97%, and 99% values were found to be lower than the control group, and size % U value was higher. In addition, the corpus callosum area was significantly lower in the ASD compared to the controls.

Conclusion

According to our study, corpus callosum of patients with ASD showed differences compared to healthy controls by histogram analysis, even though they were seen as normal in brain MRI. We think that histogram analysis can be used to evaluate possibly affected areas of brain in ASD patients.

Introduction

Autism spectrum disorder (ASD) is a neurodevelopmental disorder in which impaired social communication/interaction and stereotypical movements are seen in children.^1,2 Its prevalence is increasing worldwide. Its estimated prevalence has been reported as 1.68%. ³ Although it is stated that this increase in prevalence is due to the increase in awareness of parents and clinicians about autism, the decrease in parental prejudices, the increase in applications to physicians, and the keeping of better quality records, all these are not sufficient to explain the entire increase in prevalence. Therefore, it is very important to increase studies on the etiology of autism and to identify possible risks.

Despite all the studies carried out to date, the etiology of autism has still not been clearly determined. The generally accepted view is that it is a brain development disorder that develops as a result of interaction of multiple factors. Studies to date have focused on investigating the structural and functional alterations in brain, age of onset of these changes, and the genetic and environmental factors that cause these changes. Considering the neuroimaging studies performed on autism spectrum disorders, it is seen that there are still no specific neuropathological findings in this area. In studies conducted so far, in cases with ASD, the amygdala (social behavior and emotion), frontal cortex (attention, inhibition, and executive functions), temporal lobe (tongue), and cerebellum are the regions where the main anomalies have been reported.^4,5 It is thought that cognitive and social communication defects seen in ASD are also caused by changes in cerebral white matter pathways and deficiencies in neural connections.^6–8 The corpus callosum is the largest white matter pathway consisting of bilaterally cortical nerve fibers connecting the cerebral hemispheres. ⁹ There are some studies in the literature viewing the relationship between the pathophysiology of ASD and anomalies in the corpus callosum.^10–13

Histogram analysis is one of the texture analysis methods and has recently been widely used in the evaluation of medical images. Medical images are made up of small subunits called pixels. Each of these contains a large number of numerical data corresponding to differences that cannot be distinguished by the normal human eye. Histogram analysis provides the analysis of gray level density values at the pixel level in the image.^13,14 It is mainly used to distinguish between normal and abnormal tissues, to determine some prognostic factors, to characterize tumors, and to help the diagnosis and diagnosis process of some diseases.^15–18

In this study, we aimed to compare MRI histogram analysis values of corpus callosum of children with ASD with healthy controls.

Methods

Study population

The study is retrospective and approved by the local ethics committee of our university (03.06.2021-49709). Twenty-nine pediatric patients with ASD who underwent brain imaging between January 2016 and January 2021 and had normal brain MRIs were included in the study. For the diagnosis of ASD, all cases underwent a detailed psychiatric evaluation by an experienced child and adolescent psychiatrist, and diagnosis was made following the DSM-5 criteria. The control group included 29 age- and sex-matched healthy children with normal brain MRI and typically developing. In order for the data to be homogeneous, patients whose brain MRIs were performed with different MRI devices and those obtained with different protocols were excluded from the study. Patients with artifacts in images and pathology in brain MRI were not included in the study.

Image acquisition

Magnetic resonance imaging was obtained with a 1.5 Tesla device (Philips Medical Systems, Ingenia, Netherlands). T1-weighted 3D (TFE) turbo field echo images were obtained in the sagittal plane. Image parameters FOV: 250 mm, TR: 7 ms, TE: 3 ms, and 140 contiguous slices were used.

Image analysis

Images were uploaded to an iMac computer (27-inch). Horos Open Source Medical Image Viewer V.3.3.6 imaging software was used to measure histogram analysis. On sagittal T1-weighted images, the ROI manually bordering the corpus callosum was drawn in the midsagittal section (Figure 1). Gray level density, standard deviation, variance, entropy, skewness, kurtosis, uniformity, size %lower, size %upper, size %mean (%L; U; and M), and percentiles were measured in the ROI. A program written in MATLAB was used for image analysis (version R2017a; Natick, MA, MathWorks, USA).

Figure 1.

Sagittal T1-weighted brain MRI, ROI placement in the corpus callosum.

The gray level intensity histogram provides statistical quantitative values about the evaluated image. These are mean, standard deviation, median, minimum, maximum, variance, entropy, skewness, kurtosis, uniformity, size %L; U; and M, and percentile data.^19,20 Entropy is an indicator that the gray level intensity is not homogeneous in the measurement area. ²¹ Uniformity is a parameter that shows the uniform distribution of gray tones in the measured area. ²² The skewness shows the asymmetry in the distribution of gray tones. ²² Kurtosis refers to the peak of the distribution in gray tones. ²²

Statistical analysis

IBM SPSS 25 package program for Windows for statistical analysis was used. Mean ± standard deviation was used in expressing the data. The normality of the distribution was assessed using Kolmogorov–Smirnov test. According to the test results, Student’s t-test and Mann–Whitney U test were performed to compare the groups. For correlations using Pearson and Spearmen tests, p<0.05 value was accepted as statistical significance. ROC analysis was obtained for significant values.

Results

Of the patients with ASD, 22 (24.1%) were male and 7 (75.9%) were female. When the groups were compared in terms of gender, there was no significant difference (p = 0.620). The mean age was 7.76 ± 3.25 years in the control group and 7.69 ± 3.24 years in the patients with ASD, and no significant difference was observed (p = 0.936) (Table 1). The mean of the corpus callosum area was 3.96 ± 0.93 in the ASD group; it was 4.53 ± 0.79 in the control group and when the two groups were compared, the corpus callosum area was found to be significantly lower in the ASD group (Table 1) (Figure 2). Mean, standard deviation, maximum, median, variance, entropy, 25%, 75%, 90%, 97%, and 99% values were statistically significantly higher in the control group than in the ASD group. Size % U value was found to be statistically higher in ASD group than in the control group. There was no statistically difference between two groups in terms of other parameters (Table 2; Figure 3).

Table 1.

Statistical data of the groups according to age and corpus callosum area.

	Control (29)		ASD (29)		p
	Mean	SD	Mean	SD
Age	7.76	3.25	7.69	3.24	0.936†
Area (cm2)	4.53	0.79	3.96	0.93	0.022*

SD: standard deviation.

†Independent-samples t-test.

*Mann–Whitney U test.

Figure 2.

Distribution of the corpus callosum area values of the groups.

Table 2.

Quantitative and statistical values of histogram analysis of corpus callosum by groups.

	Control (29)		ASD (29)		p
	Mean	SD	Mean	SD
Mean	204.20	71.01	164.54	66.0	0.016*
Standard deviation	28.49	12.37	19.24	6.23	0.002*
Minimum	58.66	29.97	56.55	33.41	0.641*
Maximum	251.03	89.83	202.72	82.75	0.034*
Median	210.93	73.71	167.67	66.13	0.006*
Variance	960.19	769.74	407.83	276.92	0.002*
Entropy	5.90	0.51	5.59	0.36	0.011 †
Size % L	11.14	1.90	10.37	1.94	0.136†
Size % U	5.37	3.39	7.77	3.80	0.010*
Size % M	83.48	3.79	81.84	4.09	0.119†
Kurtosis	9.63	3.87	9.59	3.44	0.883*
Skewness	−2.23	0.55	−2.03	0.56	0.159*
Uniformity	0.59	0.06	0.56	0.07	0.106†
Percent01	90.61	34.34	88.17	47.29	0.822†
Percent03	122.29	42.57	112.42	58.04	0.463†
Percent05	141.00	51.11	126.37	62.04	0.331†
Percent10	170.92	61.56	144.41	64.17	0.101*
Percent25	199.48	69.80	159.89	63.69	0.013*
Percent75	220.26	76.78	175.52	68.76	0.023†
Percent90	228.33	79.42	182.22	70.70	0.023†
Percent95	232.52	81.22	186.05	72.21	0.015*
Percent97	235.20	82.42	188.67	73.56	0.018*
Percent99	240.82	84.81	193.69	76.38	0.029*

SD: standard deviation.

†Independent-samples t-test.

*Mann–Whitney U test.

Figure 3.

Mean (a), maximum (b), median (c), and variance (d) value distributions of groups.

When ROC analysis was obtained for the mean value (Figure 4(a)), AUC = 0.685, and the cut-off value was selected as 141.93, the ASD group could be distinguished with 65.5% sensitivity and 69% specificity.

Figure 4.

Mean (a), standard deviation (b), variance (c), and entropy (d) value receiver-operating characteristic (ROC) graphs for separating groups.

When ROC analysis was obtained for the standard deviation value (Figure 4(b)), AUC = 0.735, and the cut-off value was 22.24, the OSB group could be distinguished with 62.1% sensitivity and 72.4% specificity.

When ROC analysis was obtained for the variance value (Figure 4(c)), AUC = 0.735, and the cut-off value was selected as 398.85, the ASD group could be distinguished with 72.4% sensitivity and 65.5% specificity.

When ROC analysis was obtained for the entropy value (Figure 4(d)), AUC = 0.709, and the threshold value was chosen as 5.82, the OSB group could be distinguished with 65.5% sensitivity and 72.4% specificity.

Discussion

Autism spectrum disorders are a clinical picture that is included in childhood neurodevelopmental disorders, the symptoms begin in early childhood, and are characterized by significant inadequacies in the social-communicative area and limited, repetitive behaviors and interests. ²

The corpus callosum is the largest information transfer pathway connecting the neocortical areas of the cerebral hemispheres. ²³ In the literature, it has been reported that ASD, language disorders, intellectual disability, attention deficit, and hyperactivity disorder and learning disorders are associated with corpus callosum pathology.^24–28 Therefore, the corpus callosum has become an important field of study for researchers interested in ASD neuropathology.

Texture analysis methods are the methods that enable the determination of changes that cannot be distinguished with the normal eye and are increasing in popularity in the evaluation of medical images. Histogram analysis, which is a texture analysis method, provides numerical data of the gray level intensity of small units of images called pixels, enabling the determination of microstructural differences of tissues. ¹⁷ Some researchers have used texture analysis methods to investigate the neuropathology of some psychiatric diseases. Latha et al. heterogeneity were detected in corpus callosum, brainstem, and ventricular regions in brain MR images of patients with schizophrenia by tissue analysis methods. ²⁹ Radulescu et al. showed some differences and anomalies in gray matter and hippocampus in brain MR images of schizophrenic patients by texture analysis methods compared to healthy controls. ³⁰ In another study, using MRI histogram analysis in the corpus callosum of patients with functional neurological disorders, significant differences were shown compared to controls. ³¹

Chaddad et al. conducted radiomic analysis of subcortical brain regions on MRI in autism and found differences in tissue structure in hippocampus, choroid plexus, posterior part of corpus callosum, and cerebellar white matter in patients with ASD compared to controls. ³² According to our study, histogram analysis parameters such as mean, standard deviation, maximum, median, variance, entropy, 25%, 75%, 90%, 97%, and 99% were lower in ASD patients without any pathology in conventional brain MRI compared to control group and size %U value was higher. According to our findings, we determined that corpus callosum of ASD patients differed from healthy controls by histogram analysis. In addition, according to our study, the corpus callosum area was statistically lower in the ASD than in controls. Freitag et al. ³³ in adults with ASD, Waiter et al. ³⁴ ; Just et al. ³⁵ ; Hardan et al. ³⁶ ; and Keary et al. ³⁷ found a decrease in the size of corpus callosum in their studies on children and adults with the ASD. Prigge et al. ³⁸ found a decrease in the area of corpus callosum in children and adults. In a study, it was reported that there was no difference in corpus callosum area in the ASD patients compared to the controls. ³⁹ This study demonstrated that histogram analysis is a useful technique for demonstrating tissue differences in the corpus callosum of patients with ASD.

This study has some limitations that we are aware of. It is a single-center retrospective study and we have a small patient population. There could have been a more homogeneous age and gender group.

Conclusion

According to our research, new ideas are needed to contribute to the determination of the neuropathology of ASD. According to the findings of this study, even though the corpus callosum of patients with ASD appears normal in brain MRI, some tissue internal structure differences are shown by histogram analysis. In this way, we think that histogram analysis can be used to show the differences in the corpus callosum in ASD patients. Longitudinal assessment in pediatric patients with ASD may ultimately have an impact on the long-term prognosis and management of these patients.

ORCID iDs

Yeşim Eroğlu https://orcid.org/0000-0003-3636-4810

Murat Baykara https://orcid.org/0000-0003-2588-9013