Under Construction
The contents on this page are in the process of being created. Please check back soon.
AngioAid - A Computer-Based Platform Interpreting Coronary Angiograms
This project constructs AngioAid, a fully automated computer-based platform to assist with the interpretation of coronary angiogram videos. This project also aims to curate a dataset of coronary angiogram videos to be made publicly available via Amazon Web Services to spur development of algorithms for angiogram interpretation.
- Designed, tested, and evaluated machine learning pipeline for real-time evaluation of severity of stenosis, or thinning, of the coronary arteries.
- Cleaned and pruned large, messy data from 5 sources to create a curated data set, discarding unusable or irrelevant data
- Built and maintained an accessible data pipeline with Python and SQL
- Engineered features to extract relevant and actionable information from data regarding the severity of stenosis
- Analyzed complex model results to discover actionable insights
- Increased model performance with a novel automated data cleaning pipeline
In-Vehicle Cardiac Monitoring System
A system for detecting the onset of severe cardiac events (e.g., hemodynamic instability) prior to complications experienced by a driver.
- Designed and populated a SQL database unifying 500,000+ health records from 6+ sources into one cohesive data repository.
- Coded a pre-processing pipeline in Python to extract relevant features from text files containing clinical notes and heterogeneous patient records.
Polytrauma Decision Support System
The Polytrauma Decision Support System (DSS) technology aims to significantly improve pelvic/abdominal trauma decision-making using facilitated and prompt analysis of complex and heterogeneous patient medical data.
- Used machine learning and computer vision to design and develop real-time recommendation algorithms for clinical decision-making during time sensitive trauma treatment.
- Developed algorithms for automated segmentation of abdominal organs in CT volumes
- Trained cascading classifiers for categorization of CT volume slices based on visualized organs
- Coordinated a small group of talented graduate and undergraduate students
Privacy-Preserving Machine Learning
- Designed, implemented, and evaluated encryption-friendly adaptations of machine learing algorithms in C++
- Algorithms include privacy-preserving versions of naive Bayes, decision tree, and third-party search
- Used object-oriented programming to create a homomorphic encryption library in C++
Additional Projects
Coming Soon