Akademik Veri Yönetim Sistemi
32nd year Turkish Society of Neuroradiology congress with International Participation, İstanbul, Türkiye, 24 - 26 Mart 2023, ss.651
Objective: Vitamin D is very important for brain functions. It has been found that its deficiency causes changes such as atrophy in brain structures and is associated with cognitive and movement disorders (1). We aimed to perform texture analysis on magnetic resonance imaging (MRI) to understand the relationship between some brain structures and vitamin D in young individuals.
Materials and methods: Patients who underwent brain MRI in 3 T MRI scanners in our hospital between 2020-2022 and who had vitamin D measurements in the last 3 months were included. After the exclusion of patients with mass, stroke, and significant motion artifact, 40 patients aged 18-50 years were included. Vitamin D was low in 28 of these patients and normal in 12 patients. Of the MRI images, only T1 images without contrast were used. Sections passing through the basal ganglia level were downloaded in DICOM format and transferred to the LIFEx (version 7.3.2) texture analysis program. Here, an appropriate ROI was placed in the nucleus caudate, putamen, thalamus, and corpus callosum splenium with the consensus of 2 radiologists. For image standardization, +-3 sd was used for all images and the number of gray levels 128 was chosen. The results obtained from the texture analysis were transferred to the Orange Data mining (version 3.33.0) machine learning application. Here, the features of 40 patients were classified using the 10-fold cross-validation method. A total of 10 machine learning algorithms were used for each anatomical structure.
Results: The mean age is 35.5 in those with normal vitamin D, 30.7 in those with low vitamin D, and 32.15 in all patients. Significant relationships were observed between the texture analysis results and the vitamin D level. The most significant texture analysis parameter was NGLDM_Coarseness. The highest classification accuracy (CA) ratios for the nucleus caudate, putamen, thalamus, and splenium were 0.77, 0.70, 0.70, and 0.70, respectively to predict vitamin D level. The AUC is 0.66, 0.59, 0.67, and 0.71, respectively. The most successful machine learning algorithms for nucleus caudate, putamen, thalamus, and splenium are Tree, SVM, kNN, and SVM respectively. Considering the confusion matrix values for the nucleus caudate, which is the most successful structure, the Tree model correctly predicted 24 of 28 patients with vitamin D deficiency and 7 of 12 patients with normal values.
Conclusion: Vitamin D is of great importance for brain function and its deficiency is known to cause volumetric decreases in some brain structures such as cerebral cortex, hippocampus, and amygdala (2). In addition, it has been shown that vitamin D deficiency is associated with many neuropathological conditions such as cerebrovascular disease, dementia, Parkinson’s disease, and multiple sclerosis (3,4,5). Machine learning algorithms based on texture analysis show promise in determining the relationship between brain anatomical structures and Vitamin D levels.