Search

For many neurological disorders, including traumatic brain injury (TBI), neuroimaging information plays a crucial role determining diagnosis and prognosis. TBI is a heterogeneous disorder that can result in lasting physical, emotional and cognitive impairments. Magnetic Resonance Imaging (MRI) is a non-invasive technique that uses radio waves to reveal fine details of brain anatomy and pathology. Although MRIs are interpreted by radiologists, advances are being made in the use of deep learning for MRI interpretation. This work evaluates a deep learning model based on a residual learning convolutional neural network that predicts TBI severity from MR images. The model achieved a high sensitivity and specificity on the test sample of subjects with varying levels of TBI severity. Six outcome measures were available on TBI subjects at 6 and 12 months. Group comparisons of outcomes between subjects correctly classified by the model with subjects misclassified suggested that the neural network may be able to identify latent predictive information from the MR images not incorporated in the ground truth labels. The residual learning model shows promise in the classification of MR images from subjects with TBI.

Concussions represent a growing health concern that are challenging to diagnose and manage. Roughly four million concussions are diagnosed every year in the United States. While research in machine learning applications for concussions has focused on using advanced metrics such neuroimaging techniques and blood biomarkers, these metrics are yet to be implemented at a clinical level due to cost and reliability concerns. Therefore, concussion diagnosis is still reliant on clinical evaluations of symptoms, balance, and neurocognitive status and function. The lack of a universal threshold on these assessments makes the diagnosis process reliant on a physician’s interpretation of these assessment scores. This study aims to explore the use of machine learning techniques to aid the concussion diagnosis process. These models could provide an automated means to flag concussed patients even before being seen by a doctor as well as expand the scope of concussion diagnosis to remote locations and areas with limited access to doctors.

Clustering is a relevant exploratory tool for a broad range of machine learning applications as it aids identification of meaningful subgroups. For a given clustering algorithm, multiple partitions can be obtained on the same data set by varying algorithmic parameters. Internal validation indices provide a means to objectively evaluate how well groupings obtained from a clustering configuration partitions the data, since there is no prior labeled data. This work presents a rigorous statistical evaluation framework that analyzes performance of internal validation indices based on correlation with external indices. A synthetic data generator that captures a wide range of complexity is proposed. Evaluation is conducted on a varied set of synthetic data types and real data sets to investigate performance of the indices.

This work examines the validity of facial phenotypes as Autism Spectrum Disorders (ASD) biomarkers in boys with essential autism. A family-based association analysis framework is presented that uses previously identified facially-delineated (FD) clusters to examine relationship between FD clusters and known ASD genes. The hypothesis is that there are certain genetic variants, single nucleotide polymorphisms (SNP), specific to the FD clusters. Although statistical significance was not established, the results identified some candidate SNPs unique to each of the FD clusters that could indicate an underlying etiological difference. Further, recommendations are provided for larger-scale studies that could utilize the analysis framework presented.

Heterogeneity in Autism Spectrum Disorder (ASD) is complex including variability in behavioral phenotype as well as clinical, physiologic, and pathologic parameters. The fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) now diagnoses ASD using a 2-dimensional model based social communication deficits and fixated interests and repetitive behaviors. Sorting out heterogeneity is crucial for study of etiology, diagnosis, treatment and prognosis. In this paper, we present an ensemble model for analyzing ASD phenotypes using several machine learning techniques and a k-dimensional subspace clustering algorithm. Our ensemble also incorporates statistical methods at several stages of analysis. We apply this model to a sample of 208 probands drawn from the Simon Simplex Collection Missouri Site patients. The results provide useful evidence that is helpful in elucidating the phenotype complexity within ASD. Our model can be extended to other disorders that exhibit a diverse range of heterogeneity.