Banca de DEFESA: Lorena Batista Sandre

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : Lorena Batista Sandre
DATE: 15/12/2023
TIME: 10:00
LOCAL: Plataforma Teams
TITLE:

Radiomic signature based on cone-beam computed tomography images for evaluation of osteoporosis

KEY WORDS:

Artificial intelligence; Oral and maxillofacial radiology; Osteoporosis Cone-Beam Computed Tomography

PAGES: 100
BIG AREA: Ciências da Saúde
AREA: Odontologia
SUMMARY:

“Osteoporosis is a skeletal disease related to the loss of bone strength that predisposes individuals to fractures from minimal trauma. Fractures associated with the disease reduce quality of life for affected individuals, increase the number of hospitalizations, and can lead to increased mortality, especially in elderly, who, along with postmenopausal women, represent the highest-risk group. Additionally, fractures caused by osteoporosis are generally asymptomatic, making this disease considered silent. Therefore, methods for screening individuals at increased risk of fractures are needed to reduce socio-economic impact of the disease. Given that cone-beam computed tomography (CBCT) is a widely used examination in elderly population, especially for implant planning, and that mandibular changes have been reported in patients with osteoporosis, it becomes necessary to investigate this examination as an auxiliary tool in diagnosis. Artificial Intelligence tools based on radiomic features have been applied in diagnosis of pathologies due to their good performance in detecting tissue alterations. However, this tool has not yet been validated for detection of osteoporosis in quantitative bone analysis of the mandible by CBCT. This study aimed to determine radiomic signature of a pathological pattern in CBCT of postmenopausal women with osteoporosis. One hundred women after menopause were retrospectively selected according to criteria, who had CBCT and dual-energy X-ray absorptiometry (DXA) examination, with and without a diagnosis of osteoporosis. After determining the region of interest in CBCT, five segmentation sites were manually selected per examination: two for mandibular cortical bone, two for mandibular trabecular bone, and one for trabecular bone of axis. From segmentations, quantitative data were extracted and compiled into programming platforms, aiming to build computational codes and select the most relevant radiomic features. Forty-nine patients were diagnosed with osteoporosis according to the DXA examination, and 51 without osteoporosis. A total of 535 radiomic features per patient were extracted, distributed in seven classes: shape, firstorder, GLCM, GLDM, GLRLM, GLSZM, and NGTDM. Radiomics is a sophisticated image analysis resource, showing unique characteristics that become variables for the construction of diagnostic, prognostic, and/or predictive models. Its workflow includes various stages, including image acquisition, segmentation and feature extraction, and reproducibility. The validation of predictive radiomic models requires reproducibility and generalization of radiomic features. Strategies should be applied for building models and processing high-dimensional data that radiomics requires, as well as standardizing analysis processes.”

COMMITTEE MEMBERS:
Presidente - 2315081 - NILCE SANTOS DE MELO
Externa ao Programa - 1609346 - MYLENE CHRISTINE QUEIROZ DE FARIAS - UnBExterno à Instituição - LUCIANO FARAGE - UNIEURO
Externa à Instituição - MARIA ALVES GARCIA SANTOS SILVA - UFG