Content Details | NUML Online Research Repository

Title DETECTION OF LUNG NODULES AND CLASSIFICATION USING DEEP LEARNING NETWORK

Author(s) Salman Ahmed

Abstract One of the leading causes of cancer-related deaths around the world is lung cancer. The presence of lung nodules helps to detect lung cancer. Lung nodules are mostly small, rounded, spherical-shaped masses of vessels or tissues in the lung region. Accurate detection and classification of pulmonary nodules present in the computer tomography (CT) scan images are one of the major and complex problems. These lung nodules vary in size and shape and most of the time they are interlinked. Due to their size and location, it is difficult to detect them through naked eyes in the CT scan images. To address this problem, many researchers have used machine learning and computer vision-based techniques but these studies mostly do not consider the small size of lung nodules and ignore them. Moreover, these studies also suffer from the false-positive ratio which greatly affects the accuracy of the system. In this study, I have developed a deep learning network model VGG16 for accurate detection and classification of pulmonary nodules by considering all sizes of nodules from small sizes to large ones. Furthermore, the false-positive ratio is also improved by using unbiased data. VGG16 model stands at number one in terms of detection and it was unbeatable to date. This technique involved the steps of lung nodules image data acquisition, preprocessing of the images, data augmentation, and feature extraction using a deep convolutional neural network. After that, the deep CNN model is trained and classification of pulmonary nodules into cancerous and non-cancerous has been performed. This research work is tested and evaluated on LIDC-IDRI openly available dataset. The experimental work shows that the DCNN model VGG 16 achieved better performance with an accuracy of 93.55%, recall of 93.54%, and precision of 87.15% respectively which is better than the results gained by a previous study in this domain 91.60%

Type Thesis/Dissertation MS

Faculty Engineering and Computer Science

Department Computer Science

Language English

Publication Date 2023-01-18

Subject Computer Science

Publisher

Contributor(s)

Format

Identifier

Source

Relation

Coverage

Rights