Studente HAMAD, JAFAR
Facoltà/Dipartimento Dipartimento Ingegneria dell'Informazione
Corso di studio BIOMEDICAL ENGINEERING
Anno Accademico 2020
Data dell'esame finale 2022-02-21
Titolo italiano CLASSIFICATION OF PULMONARY CANCER HISTOTYPES FROM COMPUTED TOMOGRAPHY SCAN
Titolo inglese CLASSIFICATION OF PULMONARY CANCER HISTOTYPES FROM COMPUTED TOMOGRAPHY SCAN
Abstract in italiano Lung cancer is one of the most deadly and prevalent cancers that affect people, and it was initially designated as a global epidemic in the 1950s when many years of cigarette smoking began to inflict substantial harm. It continues to be the leading cause of cancer-related fatalities in both men and women around the world. With 2.2 million new cases diagnosed and 1.8 million deaths in 2020, lung cancer will be the second cancer type diagnosed and the first cause of cancer death, accounting for 11.4 percent of cancers diagnosed and 18.0 percent of deaths. Since 1987, lung cancer has been responsible for more deaths in women than breast cancer, and it will be the leading cause of death in men, followed by prostate and other cancers. This led to more interest in doing more research about lung cancer and to understand its symptoms, where the computed tomography played an important role by providing data related to this cancer as images, which seeks more interest in reading and analyzing the images by many researchers, especially last decade that many papers have been published related to segmentation and classification the lung cancer using neural networks and machine learning algorithms. The aim of this work is to study is to review the most important techniques to segment lung cancer in order to use later for classification. This was done by investigating literature that combined the machine learning segmentation and lung cancer and applying 3DU-net architecture with sets of scans that contain two types of cancer Adenocarcinoma and squamous. This method has been chosen as the most used and effective way to segment the CT scans and provide pure datasets for the next stage which is the classification. those kinds of literature were selected after searching on different search engines such as google scholar, PUBMED, IEEE Xplore, and 12 research were founded related to lung cancer segmentation using neural network 7 of them using U_NET. In the other hand, for classification lung cancer from CT scan techniques, and review has made by searching in 3 search engines google scholar, IEE explorer and PUBMED, that’s led us to find 8 research, 5 of them used CNN neural network. The discussed methods are used for either segmentation and classification for the lung cancer images provided from CT scans where CNN was the most used for classification, besides it was used for segmentation in some papers, where both of Men et al. paper where their results 97% sensitivity 100% specificity 94% AUC 97%, Q. Hu et al in showed accuracy 97.68%, sensitivity 96.58%, and specificity 97.11%, while Xu et al in using CNN technique for lung segmentation and their methods, while either of Shaziya et.al in 2018 and Hassani, E L and et.al in 2018, Roth et al. in 2018, Zhou et al. in 2018, and Xiao et al in 2020 used U_net Those methods applied for segmentation and classification in different ways, where LSTM and CNN were the most used techniques for image classification and sometimes researchers used a combination of both to show more efficiency especially since CNN is powerful in feature extraction with no need for pre-processing on some cases, providing more efficient results and for segmentation, it was noticeable that U_net was the most common approach used by researchers due to the feature extraction using the downsampling and upsampling processing. Machine learning and its uses in medical imaging analysis or segmentation have many approaches to follow and the experiments done by the papers above showed that the combination of two types of architecture is a better solution, providing more accurate results, making it more commonly used in the past few years.
Relatore BURATTINI, LAURA
Controrelatore DRAGONI, ALDO FRANCO
SBROLLINI, AGNESE
Appare nelle tipologie: Laurea specialistica, magistrale, ciclo unico
File in questo prodotto:
File Descrizione Dimensione Formato  
theis title_JH_signed.pdf 341 kB Adobe PDF Visualizza/Apri
thesis_jafar_hamad.pdf 2.18 MB Adobe PDF Visualizza/Apri

I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/20.500.12075/7991