Friday, September 30, 2022 - 10:00 to 12:00
SCHOOL OF COMPUTER SCIENCE
The School of Computer Science is pleased to present…
MSc Thesis Defense by: Yash Trivedi
Date: Friday September 30, 2022
Time: 10:00am to 12:00pm
Location: Essex Hall Room 122
Reminder: Two-part attendance required, Part I (QR code) and Part II (sign in sheet)
Abstract:
Lung cancer is the second most commonly diagnosed cancer and the leading cause of cancer deaths worldwide. Compared to standard chest radiography, clinical trials have demonstrated that low-dose CT significantly reduces mortality from lung cancer by 20%. However, segmenting pulmonary nodules remains challenging due to intrinsic noise, low contrast, variable sizes and different shapes of the nodules. Radiologists manually drawing the nodule boundary is a common approach for delineating nod-ules which is time-consuming, expensive and prone to intra (the difference in repeated measurements by the same observer) and inter (the difference in the measurements between observers) observer variability. Also, most lung nodule analysis techniques use supervised learning that requires manually labelled regions of interest. Therefore, an automatic unsupervised lung nodule identification algorithm is needed to assist radiologists in deciding the malignancy of nodules. The proposed method uses a set of morphological operations and median filtering in the preprocessing stage, followed by a super pixel segmentation method called Linear Spectral Clustering (LSC). After the super pixel segmentation, an eight nearest-neighbour Region Adjacency Graph (RAG) is constructed, and hierarchical agglomerative clustering is applied on the RAG in which similar super pixels are iteratively merged to form more significant regions with similar pixels. Different objects in the CT scan im-ages are clustered into unique classes. In the final step, a post-processing step of multilevel thresholding based on the colour and shape features are applied to these classes for nodule identification. The proposed method is evaluated on the Lung Image Database Consortium dataset. Comparative analysis shows that the proposed method outperforms the state-of-the-art super pixel methods for unsupervised lung nodule identification, with an average Dice Similarity Coefficient of 95.10 % and an average Intersection over Union (IoU) of 90.57 %.
Keywords: segmentation, identification super pixel, lung cancer, spectral clustering, hierarchical clustering, graph, thresholding.
MSc Thesis Committee:
Internal Reader: Dr. Boubakeur Boufama
External Reader: Dr. Esam Abdel-Raheem
Advisor: Dr. Luis Rueda
Chair: Dr. Hossein Fani
MSc Thesis Defense Announcement
5113 Lambton Tower 401 Sunset Ave. Windsor ON, N9B 3P4 (519) 253-3000 Ext. 3716 csgradinfo@uwindsor.ca