Predicting Breast Healing Deformation After Cancer Conservative Treatment

• Autor(a) principal: Pedro Miguel Martins de Lemos da Cunha Faria
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: https://hdl.handle.net/10216/106938
• topic: Engenharia electrotécnica, electrónica e informática; Electrical engineering, Electronic engineering, Information engineering

According to the annual report from the World Health Organization, breast cancer is the most frequent cancer among females. Considering all the treatments, surgery is being applied mostly using two methodologies: Mastectomy, that results on removing not only tumor, but also the total breast tissue; and Breast Cancer Conservative Treatment (BCCT) where only the tumor is removed with a thin layer of healthy tissue around it. It is clear that performing invasive treatment such as surgery, will lead to impose deformations on the breast, which can influence patients' quality of life (QoL). In this way, technology can be assisted to provide a framework that would improve the way patients interact with physicians. Enhancing this framework with the tools to visualize deformation and the healing process after the surgery can elevate patients' QoL. In order to accomplish the mentioned aim, this thesis focuses on obtaining training models to describe anatomical deformations during the healing process of the breast after BCCT. To achieve reliable training models, a dataset with several 3D breast models is required. Therefore, a semi-synthetic dataset may be generated, containing 3D breast models representing the patients' breasts before and after the surgery. The pre-surgical models are obtained through MRI data of the few patients' data that we have access. The semi-synthetic data of the pre-surgical stage will be generated taking as input these real data and variations of the hypothetic tumor's location and volume and possible breast densities. The pos-surgical data is simulated by a biomechanical wound healing model. Then by using different machine learning approaches, the relation between the patient's breast before and after the surgery can be obtained and the deformation predicted. Finally, concerning the evaluation, simulated healed breasts will be compared with the pos-surgical 3D breast models in the dataset through several metrics including Euclidean and Hausdorff distances.