Research Assistant to Dr. A.T Charlie Johnson, Rebecca W. Bushnell Professor of Physics and Astronomy at the Penn School of Arts & Sciences, and Dr. Lyle Ungar, Professor in Computer and Information Science at Penn Engineering and Psychology at the School of Arts & Sciences.

Research Overview

A team of Penn researchers led by Professor Charlie Johnson had been working on a new “electronic nose” that could help track the spread of COVID-19. The project, was recently awarded a $2 million grant from the NIH, aims to develop rapid and scalable handheld devices that could spot people with COVID-19 based on the disease’s unique odor profile. Professor Lyle Ungar is a renowned expert in the field of Artificial Intelligence and his research group actively develops explainable machine learning, deep learning, and natural language processing methods for psychology and medical research.

Being a part of the novel project, I aim to detect the presence of Covid19 disease in a human by analyzing biomarkers present as Volatile Organic Compounds (VOCs) in the sweat of a patient. The detection of VOCs is done using DNA-Functionalized Carbon Nanotube Chemical Sensors which is one of Professor Charlie Johnson’s revolutionary research outcomes.

Diseases are known to alter a number of physical processes, including body odors, and the goal of the research was to develop new ways to detect the volatile organic compounds (VOCs) that were unique to the disease. Professor Charlie Johnson’s team has developed a device with sensor arrays, an electronic version of the dog’s nose, made of carbon nanotubes interwoven with single-stranded DNA. This device binds to VOCs and can determine samples that came from patients having the disease.

My Contribution

I have been fortunate to be mentored by Professor Lyle Ungar in designing the autonomous AI disease classification system. My role in this novel project is to create autonomous, unbiased and reliable deep learning algorithm that can detect Covid19 and other medical conditions such as Ovarian Cancer, Pancreatic Cancer, Influenza, etc., by using human body odors. Each sensor from the device created by Professor Charlie Johnson’s team, measures fluctuations in current over time. This time series data from different sensors is used to classify whether the patient being tested has the disease or not.

My fully automated AI performs the following intelligent operations:

1. Capable of reading structured data from databases as well as unstructured data from text files
2. Captures all patterns generated across all sensors
3. Flags the data that’s too noisy and reports the anomalies
4. Identifies redundant sensors and level of noise in data
5. Automatically drops sensors from the analysis having high noise
6. Gives the user the flexibility of choosing specific sensors (features) for the analysis
7. Runs an ensemble of different deep learning models to report the best classification based on highest probability score
8. Appends new data from new patients to the respective database for future training
9. The AI is low latency and high efficiency