Year: 2025
Type: dissertation
Abstract: Tuberculosis (TB) remains one of the leading health concerns of the world especially in the low-resource regions where radiologists who can help one with diagnosis are scarce. Although chest X-ray imaging is prevailing examinations used in the screening of TB, it relies strongly on interpretation of experts and may prejudice performance and delay in areas that are under-privileged. Luckily, due to the advent of artificial intelligence, methods of deep learning, and Convolutional Neural Networks (CNNs) have a very high potential of automating the process of analysing medical images. The purpose of this project is to carry out the comparison of the performance of the three most popular and popular pretrained CNN architectures ResNet18, DenseNet121, and EfficientNet-B0 to see which model is more suitable in the problem of detecting TB based on a chest X-ray. We already have a publicly available dataset in Kaggle and will fine-tune the models by using transfer learning. We will use a stratified data split by training, validation and test sets to balance the set classes. All models are going to be tested in PyTorch with standard metrics such as accuracy, precision, accuracy and F1-score. The findings of the given comparison will illuminate on the trade-off between accuracy and computational efficiency of the models.
Author: Santhosh, Pooja
Institution: National College of Ireland
Language: English
Topic: COVID-19 diagnosis using AI
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7204614449