February 9, 2023

About the speaker

Mr. Moses Owusu is a graduate student in the MS Applied Statistics and Data Science program at the University of Texas Rio Grande Valley (UTRGV). He completed his Bachelor of Science in Statistics at the Kwame Nkrumah University of Science and Technology (KNUST), Ghana, where he graduated with a first-class honour. He is hardworking, dedicated and eager to learn. 

After his BSc, he served as a teaching and research assistant at the department of Mathematics, KNUST where taught several undergraduate courses and assisted students with their research in areas such as time series and survival analysis. His research interests include Finance, Disease modelling, Operations Research, and Data Science. 

In summer 2022, he worked as a computational science graduate intern in the Theoretical Biology and Biophysics Group (T-6) at the Los Alamos National Laboratory (LANL), New Mexico. There, he worked on projects including mosquito-borne disease modelling specifically, West Nile virus and dengue fever. Contributing through spatial sampling of ebird data as well as preparing geological data from Brazil for analysis that involved multi-host vectors, human demographics, and climate data. 

He is currently working under the supervision of Dr. Hansapani P. Rodrigo to use Electrocardiogram (ECG) readings to identify and classify different heart disorders with the use of Symbolic Aggregate Approximation (SAX). After his MS degree, he plans to further with a PhD in Biostatistics and contribute through collaborative and pioneering research in Health, impart knowledge in statistics as a multi-lingual professor, while travelling the world.

Identification of heart diseases using Symbolic Aggregate Approximation (SAX)

This project is an application of the Symbolic Aggregate Approximation (SAX) to 1000 fragments of ECG signals for 45 patients (42% females aged between 23 and 89 years and 58% males aged 32 to 89 years) using data obtained from the MIH-BIH Arrhythmia database to recognize cardiac health disorders. Data include a normal sinus rhythm, pacemaker rhythm and ECG readings for 15 heart disorders, making 17 in total. The aim is to use SAX to classify heart disorders using ECG signal, that analyzes QRS-complexes by first splitting the time series into smaller equally sized segments using the Piecewise Aggregate Approximation (PAA) approach. SAX is then used to reduce the normalized PAA series to a string of arbitrary length alphabets. Using a sliding window algorithm, we create a list of word bags that are the result of SAX on our data set – this gives the words as well as frequency counts. We then specify the PAA size, window size and SAX alphabet size, and compute weights for these word bags that give an indication of the proportion of these words sequence in our input time series. This will be done to the sample data set for each heart disorder and at the end, we are able to determine which exact patterns contributes the most (motifs) and the least (discords) to our class of words, facilitating the identification of possible heart disorders.