Research Interests:
Deep Learning, Medical Image Analysis and Cognitive Modeling.

We are currently working on the optimization of deep learning structure and its application in task engagement assessment, speech signal processing for Posttraumatic stress disorder (PTSD) patients diagnosis and medical image analysis.

PTSD patients diagnosis: PTSD is an anxiety disorder that can develop after exposure to one or more traumatic events that threatened or caused great physical harm. It is often difficult and costly to diagnose PTSD due to its varieties of symptoms and potential patients are often reluctant to seek help due to the stigma associated with it. Such a difficulty can be mitigated if a low cost and non-intrusive tool can be developed to remotely monitor the mental healthiness of the war fighters without a lengthy interview conducted by clinicians. Past research has shown several prominent features of speaking behavior and voice sound characteristics to be closely related to the severity of patients’ mental illness as well as the time course of recovery from depression. We propose to develop an advanced deep learning model to automatically extract salient features from audio. Those features will be further fused with previously proved handcrafted audio features to boost PTSD diagnosis. In addition, ODU will assist IAI on software testing and simulation.

Remote sensing image processing: We developed a sparse coding based dense feature representation model for hyperspectral image (HSI) classification. We integrated the method with a linear support vector machine (SVM) and a composite kernels SVM (CKSVM) to discriminate different types of land cover. We evaluated the proposed algorithm on three well known HSI datasets and compared our method to four recently developed classification methods including SVM, CKSVM, simultaneous orthogonal matching pursuit (SOMP) and image fusion and recursive filtering (IFRF). Experimental results show that the proposed method can achieve better classification accuracies with a much more compact representation leading to more efficient sparse models for HSI classification.  We are now investigating if the deep convolutional neural network (CNN) can bring extra advantages in the land cover identification task.

Engagement assessment: We study and improve task engagement assessment of pilots for aviation safety. The major work addresses two challenging problems involved in the assessment: individual variation among pilots and the lack of labeled data for training assessment models. Task engagement is usually assessed by analyzing physiological measurements collected from subjects who are performing a task. However, physiological measurements such as Electroencephalography (EEG) vary from subject to subject. An assessment model trained for one subject may not be applicable to other subjects. We proposed a dynamic classifier selection algorithm for model individualization to address this challenge. For complex tasks such as piloting an air plane, labeling engagement levels for pilots is difficult. We proposed to utilize deep learning models to tackle this lack of labeling information issue.

Deep learning for AD patients classification: Accurate classification of Alzheimer’s Disease (AD) and its prodromal stage, Mild Cognitive Impairment (MCI), plays a critical role in possibly preventing progression of memory impairment and improving quality of life for AD patients. Among many research tasks, it is of particular interest to identify noninvasive imaging biomarkers for AD diagnosis. We present a robust deep learning system to identify different progression stages of AD patients based on MRI and PET scans. We utilized the dropout technique to improve classical deep learning by preventing its weight co-adaptation, which is a typical cause of over-fitting in deep learning. In addition, we incorporated stability selection, an adaptive learning factor, and a multi-task learning strategy into the deep learning framework.