Texture Analysis of Corpus Callosum in Mild Traumatic Brain Injury Patients

Texture analysis (TA) is a quantitative approach for characterizing subtle changes in magnetic resonance (MR) images of different tissues. The aim of this study was to detect changes in tissue of corpus callosum (CC) in mild traumatic brain injury (MTBI) patients by the means of TA. TA was performed in the sagittal T1-weighted MR images of 42 MTBI patients, focusing on different segments of CC by using the tissue characterization software MaZda. Results were compared with the control group of ten healthy volunteers. The most discriminant texture features were identified with a combination of feature selection algorithms mutual information (MI), classification error probability combined with average correlation coefficients (POE+ACC) and Fisher coefficient. Linear discriminant analysis (LDA) and nonlinear discriminant analysis (NDA) were performed. Nearest-neighbor (1-NN) classification for LDA and artificial neural network (ANN) for NDA was used for tissue classification. The results revealed differences in the textures between the selected segments of CC in MTBI patients. There were also differences in the CC between healthy volunteers and MTBI patients. The best classification results between healthy volunteers and patients were achieved in the area of splenium of CC, with accuracy of 96% for the 1-NN classifier, and accuracy of 98 % for the ANN classifier. TA results revealed changes in the texture parameters of the segments of CC between healthy volunteers and MTBI patients and therefore may provide a novel additional tool for detecting subtle changes in CC tissue on MTBI, but evidently larger data is necessary to confirm the clinical value of TA in diagnosing MTBI.