Research Article
Identify Tumors on Lung CT Images
@INPROCEEDINGS{10.1007/978-3-031-58878-5_12, author={Phong Thanh Le and Thai Hoang Le and Hieu Duc Thai Tran}, title={Identify Tumors on Lung CT Images}, proceedings={Context-Aware Systems and Applications. 12th EAI International Conference, ICCASA 2023, Ho Chi Minh City, Vietnam, October 26-27, 2023, Proceedings}, proceedings_a={ICCASA}, year={2024}, month={8}, keywords={You Only Look Once Tumor identification Lung Computed Tomography (CT) images Lung Nodule Analysis 2016 (LUNA16) dataset Generalized Intersection over Union (GIoU) loss Diagnosis Lung cancer}, doi={10.1007/978-3-031-58878-5_12} }- Phong Thanh Le
Thai Hoang Le
Hieu Duc Thai Tran
Year: 2024
Identify Tumors on Lung CT Images
ICCASA
Springer
DOI: 10.1007/978-3-031-58878-5_12
Abstract
This paper introduces You Only Look Once (YOLO) model to identify tumors on computed tomography (CT) lung images. The model uses a variant of the YOLO algorithm called YOLOv5 [1], which is known for its accuracy and speed in object detection tasks. To train and evaluate the YOLO model, we use the Lung Nodule Analysis 2016 dataset (LUNA16) [2]. This dataset contains a set of lung CT scans with annotations indicating the location of the tumors. We preprocess the CT images and annotations to prepare data for model training and testing. During the training phase, the YOLO model uses a loss function named Generalized Intersection over Union (GIoU) loss [3], which provides a more accurate measure of box overlap between the predicted objects and the ground truth. The combination of the YOLO architecture and the GIoU loss enables accurate and fast detection, making the proposed model a promising tool to aid physicians in diagnosing lung cancer.
