The research goal of our laboratory is to study cutting-edge computational and informatics methods, and turn them into practical tools that can help computers better understand digital data (e.g., 1-D sequences, 2-D images, and 3-D shapes) in practical applications. In particular, we are recently working on various neuroimaging and bioinformatics projects. In these projects, we apply, extend or develop state-of-art software tools for analyzing structural and functional neuroimaging data as well as genomic and other related biomarker data. The ultimate goal is to improve early diagnosis and mechanistic understanding of disease processes and treatment response for brain disorders such as Alzheimer's disease and schizophrenia, as well as the cognitive effects of cancer chemotherapy. We are also involved in two Fetal Alcohol Spectrum Disorders (FASD) studies, where we develop facial imaging methods to identify features important in FASD as well as examine a parallel mouse model to examine how timing of alcohol exposure influences the pattern of facial dysmorphology. The computational methods we study in these biomedical applications include (1) image & vision computing, (2) data mining & pattern recognition, and (3) geometric modeling & graphics.

This research program is intensively multidisciplinary. We have assembled an excellent set of critical resources, including (1) experts from neuroscience, neuroimaging, computer science, genetics, informatics, statistics, and related domains, (2) state-of-the-art imaging facilities, and (3) state-of-the-art computer systems (access to supercomputers at IU, powerful workstations, file & web servers) and software tools. These resources greatly enable our capability to not only collect large volumes of imaging and other biomarker data but also study major computational and analytic challenges limiting the effective application of these data and tools towards solving complex biomedical problems.

Our previous computational work is mostly focused on medical image analysis and bioinformatics. We have been developing computational methods for various challenging problems along the processing pipeline of many medical imaging and genetics studies. Below are a few examples.

(1) Image Segmentation and 3D Modeling: How to extract features from 3D medical images (e.g., MRI, fMRI, DTI)? For example, we have developed automated methods for extracting pulmonary nodules from lung images. We have also developed surface modeling methods to characterize the shape of certain regions of interest (ROIs).

(2) Image Registration and Shape Alignment: How to make these features comparable across different subjects or over the time dimension so that group analysis or temporal analysis can be facilitated? In this area, we have been studying registration algorithms to align images/objects together (e.g., the same lung nodule over different time points, MRI scans or ROIs across different subjects, cross-modality registration of the same subject (e.g., fMRI to MRI, DTI to MRI)).

(3) Pattern Analysis and Knowledge Discovery: How to use these comparable features to perform statistical analysis and pattern recognition? Examples of our studies in the area include: (1) using hippocampal shape information to predict schizophrenia (shape classification), (2) localizing hippocampal shape changes in MCI/AD (statistical morphometric analysis), and (3) relating imaging phenotypes to interesting genotypes (imaging genetics).

Personalized Therapeutics Group
TRIP Translating Research into Practice