Democritus University of Thrace (DUTH)
Machine learning methodologies for breast cancer diagnosis from mammograms
Abstract
dc:descriptionThis thesis aims to develop innovative and effective computer-aided diagnosis methodologies, regarding breast cancer lesions in mammograms. Specifically, after an extensive analysis of state of the art works, two methodologies are proposed.In the first one, the incorporation of content-based image retrieval (CBIR) into computer aided diagnosis (CADx) is investigated. In this way, decision is provided in the context of similar retrieved cases, in order to contribute to the decision-making process of radiologists in the characterization of mammographic masses. The proposed scheme comprises two stages: a margin-oriented supervised CBIR stage that retrieves images from reference cases along with a decision stage that is based on the retrieved items. The feature set utilized exploits state-of-the-art features along with a newly proposed texture descriptor, namely mHOG, targeted to capturing margin and core specific mass properties. Performance evaluation considers the CBIR and diagnosis stages separately and is addressed by using standard measures on an enhanced version of the widely adopted digital database for screening mammography (DDSM). The proposed scheme achieved improved performance of CADx of masses in X-ray mammography experimentally compared to the state-of-the-art.The second approach follows a deep learning classification pipeline. The network architecture and training type is selected after a detailed study, where different state-of-the-art convolutional neural networks are examined in two training scenarios: the first involves initializing the network with pre-trained weights, while for the second the networks are initialized in a random fashion. Extensive experimental results show the superior performance achieved in the case of fine-tuning ResNet-50. Based on that, the main contribution of the pipeline is a novel methodology for integrating segmentation information into a CNN, aiming to improve breast cancer diagnosis. The proposed methodology involves modification of each convolutional layer of a CNN, so that information of not only the input image but also the corresponding segmentation map is considered. Furthermore, a new loss function is introduced, which adds an extra term to the standard cross-entropy, aiming to steer the attention of the network to the mass region, penalizing strong feature activations based on their location. The segmentation maps are provided either by domain experts using a semi-automatic procedure or from an automatic segmentation stage. The experimental results demonstrate that the integration of segmentation in the proposed scheme improved breast cancer diagnosis performance compared to state-of-the-art methods.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Tsochatzidis, Lazaros
- Τσοχατζίδης, Λάζαρος
Subjects
dc:subject × 24- Μαστογραφία
- Ανάκτηση βάσει περιεχομένου
- Βαθειά μάθηση
- Αλγόριθμοι ταξινόμησης
- Εικόνες, τμηματοποίηση
- Mammography
- Content-based image and video retrieval
- Mammography screening
- Deep learning
- Support vector machines
- Computer aided diagnosis
- Segmentation
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Τεχνολογία μέσων
- Επιστήμες Υγείας
- Ιατρική και Επιστήμες Υγείας
- Ιατρική Πληροφορική
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Media Technology
- Health Sciences
- Medical and Health Sciences
- Medical Informatics
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/49767