Abstract:Background. Mild traumatic brain injury (mTBI) is a widespread and serious public health problem which also causes physical and psychological suffering to patients and their families and imposes a significant economic burden on society. But it is usually very difficult to detect and provide warning of mTBI in early stage. Therefore, a novel method is urgent for the increasing demands on the accurate and rapid prediction and feature selection of mTBI. Objectives. To establish a better idea of the performance of neuroimage biomarker in the acute phase of mTBI, our study adopts diffusion tensor imaging (DTI) which could present the pathophysiological changes of white matter through several parameters noninvasively and combined with behavioral experiments such as intelligence quotient test, memory, executive function, and motion function to find the relationship between DTI abnormal brain regions and behavioral abnormalities. Then, provide new method for rapid prediction and feature selection of mTBI. Methods. 77 mTBI patients were admitted to the Emergency and Neurosurgery Departments of the Third Xiangya Hospital of Central South University from August 2019 to July 2021; the patients (41 males and 36 females) suffered mTBI because of car accident (36), assault (11), and fall (30). All the mTBI patients were examined through MRI scan and behavioral psychology test within 3 days after injury. MRI images and behavioral psychology tests were also collected; the correlation between the DTI biomarker and the cognitive psychological outcome was analyzed. A series of integration and computational methods were also used for fusion arithmetic and result analysis. Results. Compared with the healthy control group, the patients in the acute stage of mTBI presented lower scores in the digit symbol substitution test (DSST), suggesting that mTBI patients in the acute stage had decline in information processing speed and associative learning. The difference of DTI parameters in acute stage mTBI patients was mainly manifested as increased AD and MD values in multiple brain regions, while RD and FA values have no significant difference. The most significant brain regions were bilateral corticospinal tracts (CST), bilateral posterior internal capsule lentiform nucleus, bilateral superior longitudinal fasciculus, left terminal striae, and left sagittal plane with right posterior thalamic radiation. The Pearson correlation coefficient was significantly positive correlation between AD and MD elevation in the left sagittal layer and the results of DSST and digit span in acute stage mTBI patients. Conclusions. The acute phase mTBI patients performed lower score on the DSST than those in the normal control group. This neuropsychological change was associated with increased AD value and MD value in the left sagittal layer, which indicated reduction of information processing speed in mTBI patients in the acute phase. It might be related to abnormal AD value and MD value in the upper longitudinal tract, lower longitudinal tract, lower frontal occipital tract, and sagittal layer. In this study, combined with neuropsychological test and increase of the AD value and MD value in certain brain region, neurosurgeon should pay more attention to the abnormal of the upper longitudinal tract and the patients' information processing speed in the diagnosis and treatment of the acute phase mTBI patients. The study offers a much more secure and integrated method for rapid prediction and feature selection of mTBI, which could have broader clinical approaches and application prospects.

A Deep Unsupervised Learning Approach Toward MTBI Identification Using Diffusion MRI

MTBI Identification From Diffusion MR Images Using Bag of Adversarial Visual Features

A Machine Learning Approach For Identifying Patients with Mild Traumatic Brain Injury Using Diffusion MRI Modeling

Identifying Mild Traumatic Brain Injury Patients From MR Images Using Bag of Visual Words

Deep learning-based multimodality classification of chronic mild traumatic brain injury using resting-state functional MRI and PET imaging

Identification of Relevant Diffusion MRI Metrics Impacting Cognitive Functions Using a Novel Feature Selection Method

Detection of Chronic Blast-Related Mild Traumatic Brain Injury with Diffusion Tensor Imaging and Support Vector Machines

Validation of a Deep Learning Model for Traumatic Brain Injury Detection and NIRIS Grading on Non-Contrast CT: a Multi-Reader Study with Promising Results and Opportunities for Improvement.

Identification of chronic mild traumatic brain injury using resting state functional MRI and machine learning techniques

Subject-Specific Abnormal Region Detection in Traumatic Brain Injury Using Sparse Model Selection on High Dimensional Diffusion Data

Rapid Prediction and Accurate Location Selection of Mild Traumatic Brain Injury (mTBI) by Using Multiple Parameter Analysis of Diffusion Tensor Imaging (DTI): Integrating Correlational and Clinical Approaches

A Siamese Convolutional Neural Network for Identifying Mild Traumatic Brain Injury and Predicting Recovery

Machine Learning Applications in Traumatic Brain Injury: A Spotlight on Mild TBI

Interpretable 3D Multi-Modal Residual Convolutional Neural Network for Mild Traumatic Brain Injury Diagnosis

Improved Traumatic Brain Injury Classification Approach Based on Deep Learning

Discriminating mild traumatic brain injury using sparse dictionary learning of functional network dynamics

Three dimensional convolutional neural network-based automated detection of midline shift in traumatic brain injury cases from head computed tomography scans

Diffusion Model based Semi-supervised Learning on Brain Hemorrhage Images for Efficient Midline Shift Quantification

Toward more accurate and generalizable brain deformation estimators for traumatic brain injury detection with unsupervised domain adaptation

Brain Age Prediction Using Interpretable Multifeature-based Convolutional Neural Network in Mild Traumatic Brain Injury

Prognosticating outcome using magnetic resonance imaging in patients with moderate to severe traumatic brain injury: a machine learning approach