ELLIE C CHEN


MACHINE LEARNING-BASED EARLY DETECTION SYSTEM



PROJECT OVERVIEW





TO DETECT DISEASES EARLIER IN THEIR STAGES through the use of risk factors as features in a ml algorithm



The basis behind this project is applicable to countless diseases, however, this project, specifically, was made in hopes of increasing the number of detected TB cases in Ghana. A trip to the country itself proved hospitals' accessibility to GeneXpert, which is able to detect TB in less than 2 weeks from exposure. However, the current health system only refers patients to get TB tested during admission as an HIV patient and display of coughing symptoms, which generally appears in the third month of contraction. The use of risk factors (such as but not limited to proximity to diagnosed TB patients, malnutrition, and immunodeficiency), rather than simply symptom-based diagnosis and referrals would increase the number of cases referred to TB clinics for testing.



1



Heart Disease Proof of Concept





Having started the project months before the trip to Ghana, we used heart disease data, retrieved from Kaggle, to create a desktop application that would input features such as age, resting blood pressure, and presence of angina and output the likelihood of a patient having heart disease.



2



Application to TB detection





The questions asked to the right are potential questions for the TB application of this program, designed after much discussion with physicians in Ghana.



3



Future Direction





The current status of the project is in the talks with physicians in Ghana over data collection using the features we have selected for our algorithm. We will then use this data to train our program so it can be implemented in clinics such as Suntreso and Kumasi South that we have been talking to.



TIMELINE AND SKILL SETS





Spring 2019 - Present

Java/Netbeans

Python

Machine Learning (Bootstrapping)

Design



RELEVANT LINKS & DOCUMENTS




PROJECT WEBSITE