Průzkum a porovnání metod výběru dat pro předzpracování vstupu modelu hlubokého učení pro automatickou segmentaci nádorů na snímcích PET-CT
Exploring and comparing data selection methods in the pre-processing step of a deep learning framework for automatic tumor segmentation on PET-CT images
Typ dokumentu
diplomová prácemaster thesis
Autor
Ekaterina Korotina
Vedoucí práce
Ráfl Jakub
Oponent práce
Reimer Michal
Studijní program
Biomedicínské a klinické inženýrstvíInstituce přidělující hodnost
katedra biomedicínské technikyPráva
A university thesis is a work protected by the Copyright Act. Extracts, copies and transcripts of the thesis are allowed for personal use only and at one?s own expense. The use of thesis should be in compliance with the Copyright Act http://www.mkcr.cz/assets/autorske-pravo/01-3982006.pdf and the citation ethics http://knihovny.cvut.cz/vychova/vskp.htmlVysokoškolská závěrečná práce je dílo chráněné autorským zákonem. Je možné pořizovat z něj na své náklady a pro svoji osobní potřebu výpisy, opisy a rozmnoženiny. Jeho využití musí být v souladu s autorským zákonem http://www.mkcr.cz/assets/autorske-pravo/01-3982006.pdf a citační etikou http://knihovny.cvut.cz/vychova/vskp.html
Metadata
Zobrazit celý záznamAbstrakt
Automatic segmentation of primary tumors in oropharyngeal cancer patients using PET/CT images and deep learning has the potential to improve radiation oncology workflows. However, 2D tumor segmentation using deep learning is a data imbalance problem and a method of PET and CT slice selection affects the convergence of the deep learning model. The aim of the current project was to find a way to select sequences to improve the performance of the existing deep learning segmentation model. To select the 'right amount' of sequences without tumor in an unsupervised manner, clustering methods were explored. The trained clustering algorithms were used to group the training and validation data of the existing segmentation model in into clusters. The performance of the proposed method was assessed using the existing segmentation model. The promising results of the proposed data selection method were confirmed by improved metrics of the segmentation model (mean dice score coefficient, precision and recall). Automatic segmentation of primary tumors in oropharyngeal cancer patients using PET/CT images and deep learning has the potential to improve radiation oncology workflows. However, 2D tumor segmentation using deep learning is a data imbalance problem and a method of PET and CT slice selection affects the convergence of the deep learning model. The aim of the current project was to find a way to select sequences to improve the performance of the existing deep learning segmentation model. To select the 'right amount' of sequences without tumor in an unsupervised manner, clustering methods were explored. The trained clustering algorithms were used to group the training and validation data of the existing segmentation model in into clusters. The performance of the proposed method was assessed using the existing segmentation model. The promising results of the proposed data selection method were confirmed by improved metrics of the segmentation model (mean dice score coefficient, precision and recall).
Kolekce
- Diplomové práce - 17110 [1011]