Tuberculosis Detection Using Pre-Trained CNNs

Kothiyal, Ekansh (2024) Tuberculosis Detection Using Pre-Trained CNNs. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (5MB) | Preview
Preview	PDF (Configuration Manual) Download (923kB) | Preview

Abstract

Tuberculosis (TB) is a leading infectious disease globally, that continues to emphasize its high mortality rate and contagious properties, especially in low and middle-income regions. There have been many diagnostics in this area. Still, with current diagnostics methods, they face huge limitations in sensitivity, speed, and accessibility which leads to delays in early TB detection and treatment. Deep learning models are explored in this study, specifically convolutional neural networks (CNNs), which act as a tool to enhance TB diagnostics that help classify all the cell images as TB-infected or healthy cells. Two pre-trained Models MobileNetV2 and DenseNet121, have been investigated for their effectiveness in TB image classification. In this study, MobileNetV2 is selected due to its computational efficiency and DenseNet121 is chosen due to its deep feature extraction capabilities. The models were trained using transfer learning on a sputum smear microscopy dataset with TB-infected and healthy cell images. The findings in this research highlight the viability of pre-trained CNN models as an important tool in TB diagnosis.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Subhnil, Shubham UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science R Medicine > Diseases R Medicine > Healthcare Industry Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	03 Sep 2025 10:58
Last Modified:	03 Sep 2025 10:58
URI:	https://norma.ncirl.ie/id/eprint/8731

Actions (login required)

View Item