Mild traumatic brain injury (MTBI) affects ~1.5 million persons annually in the United States with fifteen to 30% of patients suffering long-term disability after injury. We remain in the early phase of understanding this disease and one of the greatest barriers to studying the disease and developing appropriate therapy is the difficulty in diagnosis and outcome prediction. Generally, the diagnosis of MTBI relies on using the Glasgow Coma Scale (GCS), a 15-point gross measurement of eye-opening, motor and verbal response. The National Institute for Neurological Disorders and Stroke (NINDS) workshop in 2014 indicated that use of GCS score as a single classifier for TBI is insufficient and proposed that neuroimaging play a larger role towards the development of objective criteria for diagnosis and outcome prediction. We have specific experience in studying novel MRI techniques that show much promise in evaluating MTBI patients. The goal of the current proposal is to bring these novel MRI techniques to clinical use. We propose to combine information from objective MR imaging features with clinical information to learn the patterns that can best distinguish patients from controls and predict long-term outcome using machine learning. We will validate our tool using a separate subject cohort. Such a tool would be an extremely powerful clinical tool to identify at-risk patients for early intervention. Additionally, this research will identify the most clinically relevant MR metrics, thereby pointing the way to novel therapeutic pathways.
Mild traumatic brain injury (MTBI) is a major public health problem for which there is a lack of evidence-based, quantitative and objective criteria for diagnosis and outcome prediction. The goal of the proposed research is to incorporate recent advances in MR imaging of MTBI with clinically important information using advanced machine-learning computational algorithms to identify the most clinically relevant features, thus allowing us to distinguish patients from controls and to predict clinical outcome. If successful, this research will provide an objective tool for classification and outcome prediction in MTBI and will be a critical advance in both the clinical and research arenas in the study of traumatic brain injury.
| Wang, Yuan; Wang, Yao; Lui, Yvonne W (2018) Generalized Recurrent Neural Network accommodating Dynamic Causal Modeling for functional MRI analysis. Neuroimage 178:385-402 |
| Lui, Yvonne W; Xue, Yuanyi; Kenul, Damon et al. (2014) Classification algorithms using multiple MRI features in mild traumatic brain injury. Neurology 83:1235-40 |
