Date of Award
Santa Clara : Santa Clara University, 2022.
Computer Science and Engineering
One of the top 10 causes of death, 1.5 million people die from tuberculosis (TB) every year. And yet, in 2018, 1.7 billion people were infected with the disease (“Global Health - Newsroom - Tuberculosis.”). This 99.99% recovery rate demonstrates that, when treated with antibiotics, TB is an extremely curable condition, but an early diagnosis is critical to surviving the illness. Developing countries face the largest hurdles to accessing TB screening technologies, and they make up a disproportionately high fraction of worldwide TB deaths. A possible solution to this problem would be to use machine learning (ML) to detect TB on cough sounds. However, there currently is not enough data to develop such an application. In this paper, we discuss our android tablet application (Ekifuba Test) that begins the process of collecting cough and demographics data across Uganda. This data will be used to accrue training data for a future TB diagnosing ML application. By eliminating the need to carry around medical equipment, replaced with a single android tablet, this ML model would serve as a vastly more accessible alternative to our contemporary TB diagnostic approaches. Currently in the testing stage, we have delivered our data collection application to Ugandan clinicians, who will shortly begin using our technology in the field to accrue real-world samples. Once we have collected sufficient data, two of our team members will work to develop the aforementioned ML model by training it on the acquired samples–in an effort to finally create an affordable, accessible solution to the TB diagnosing problem.
Vinolus, JT; De La Torre, Daniela; Kerr, Aimee; and Randleman, Jordan, "Large Scale Database Audio Collection of Tuberculosis Patients: : the Ekifuba Test" (2022). Computer Science and Engineering Senior Theses. 234.