Banca de QUALIFICAÇÃO: ANTHONY EMANOEL DE ALBUQUERQUE JATOBA

Uma banca de QUALIFICAÇÃO DE MESTRADO foi cadastrada pelo programa.

DISCENTE: ANTHONY EMANOEL DE ALBUQUERQUE JATOBA
DATA: 03/08/2020
HORA: 10:00
LOCAL: Video-Conferência
TÍTULO:

DEEP LEARNING-BASED LUNG NODULE CLASSIFICATION ON MULTIMODALMEDICAL IMAGING

RESUMO:

Lung cancer is the most common and lethal form of cancer, and its early diagnosis is key to the patient’s treatment and survival. Early detection can be achieved through screening programs, whose gold-standard scan is the Computed tomography (CT). However, this imaging modality presents the drawback of exposing the patient to ionizing radiation. Recent studies have shown the relevance of magnetic resonance imaging (MRI) in lung cancer assessment and its complementary role to CT in the detection and characterization of pulmonary nodules, with no need for radiation exposure. Technological advancements also allowed for the use of multimodality medical imaging in clinical practice and research, aiming to take advantage of the complementary information present in different images modalities.

In this proposal, we present the current state of multimodality image systems for lung cancer assessment and how we intend to contribute to this research area by investigating whether multimodality CT/MRI deep learning-based models for lung nodule classification can provide better performance compared to its single modalities counterparts; comparing different strategies to fuse the images information into a predictive model; and assessing the contribution of individual modalities to the overall performance.

Our methodology to reach those objectives is presented, encompassing an image registration protocol for aligning CT and MRI images; a semi-automatic nodule segmentation protocol; the architecture for three convolution neural networks, corresponding to three image fusion strategies; our model development general guidelines; and our performance evaluation and validation strategy.

Lastly, we present preliminary results on multimodality radiomics-based models and the planned activities and schedule.

PALAVRAS-CHAVE:

Multimodal Imaging; Computed tomography (CT); Magnetic Resonance Imaging (MRI); Deep Learning; Convolutional Neural Network (CNN); Lung Cancer; 

PÁGINAS: 40
GRANDE ÁREA: Ciências Exatas e da Terra
ÁREA: Ciência da Computação
SUBÁREA: Metodologia e Técnicas da Computação
ESPECIALIDADE: Processamento Gráfico (Graphics)

MEMBROS DA BANCA:
Presidente - 1544992 - MARCELO COSTA OLIVEIRA
Interno(a) - 1766576 - THALES MIRANDA DE ALMEIDA VIEIRA
Externo(a) à Instituição - PAULO MAZZONCINI DE AZEVEDO MARQUES - USP