Research

Deep Learning

Disease Prediction, Drug Discovery, Depression detection, Word Embedding, Molecule Representation, Graph Learning

Deep Adversarial Domain Mapping

Deep Learning for Depression Detection using Smartphones and Wearable Devices

DeepChem: Drug Discovery and Molecule Representation

Contextual Embedding of Medical Concepts from Electronic Health Records

High-Performance Computing Framework on GPU to Accelerate Deep Learning

Data Mining for Healthcare Data

EHRs, Medical Image (MRI), Clinical Data, Genetic Data, Smartphone, Fitbit etc.

Generative Medical Image Modeling with Generative Adversarial Networks

• Learn the joint distribution of multi-domain medical images.
• Generate pairs of corresponding medical images.

LifeRhythm: Automatic and Pervasive Depression Screening Using Smartphones and Wearable Devices

• Develop an automated system for automatic and pervasive depression screening using smartphone data.
• Monitor the behavioral rhythms of individuals through their smartphones, extracts normalized features from the raw data, and applies multiple machine learning models for real-time diagnosis.
• Joint Modeling of Heterogeneous Sensing Data for Depression Assessment via Multi-task Learning.

Contextual Embedding of Medical Concepts from patient EHRs

• Construct contextual embedding of medical concepts from patient EHRs with Word2vec technique.
• Investigate multi-sense mechanism to allow same medical concept with different context during the course of disease progression.

Computational Phenotyping

• Investigate existing techniques for computational phenotyping.
• Evaluate patient similarities based on the extracted phenotypes.
• Privacy issues on computational phenotyping.

Visual Analytics for Healthcare Data

• Extract features from patient EHR.
• Visualize the longitudinal EHR of a patient cohort with Sankey Diagram type of techniques.

Unstructured Data Analytics

Text Mining

Information Extraction from Unstructured Text Data

• Research effective Data Mining and Knowledge Discovery techniques for emerging unstructured data.
• Learn sentence patterns and calculate the semantic similarity between the text and set of patterns.
• Extract most relevant terms and its associated parameters from unstructured big data.