Research projects

Disease Risk Prediction

This project involved leveraging routine EHR data to develop AI risk prediction models. Models for personalized prediction of risk chronic diseases (such as heart failure, COPD, pneumonia, diabetes, atrial fibrillation, chronic kidney disease), hospital readmissions, and mortality were developed. The AI models outperformed existing validated clinical models such as; LACE for predicting hospital readmissions; GWTG-HF for mortality, and so on. Our cardiovascular risk prediction model was among top models in the PrecisionFDA challenge.

Automatic Data Extraction to Support Meta-Analysis Statistical Analysis

This project gives an overview of my Ph.D. research on contributing to automation of meta-analysis. Meta-analysis is the process of collecting and analyzing results of different studies that are focussed on same treatment or disease to ascertain if a treatment is effective or not. Meta-analysis provide the gold standard for medical evidence. However, despite their importance, meta-analyses are time- and cost-consuming and this is challenging especially in cases where timeliness is important. In this research, I am developing a system for automating the meta-analysis process, especially the data extraction and statistical analysis steps.

READ MORE

Contextualized medication event extraction

This project describes participation in the n2c2 2022 shared task, Track 1 on Contextalized Medication Event Extraction (CMED). The task is to classify medication mentions into three categories based on the context.

  1. Dispostion - there was medication change discussed
  2. NoDisposition - there was no medication change discussed
  3. Undetermined - unclear if a change was discussed, more information is neeeded

READ MORE

Semantic textual similarity in clinical domain texts

Semantic Textual Similarity (STS) captures the degree of semantic similarity between texts. STS plays an important role in many natural language processing applications such as text summarization, question answering, machine translation, information retrieval, dialog systems, plagiarism detection, and query ranking. In this research I discuss BERT-based approaches to calculate semantic similarity in English and Japanese clinical texts.

READ MORE