• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Doctoral Thesis
DOI
https://doi.org/10.11606/T.17.2021.tde-11042022-133150
Document
Author
Full name
Hélcio Mendonça Pereira
Institute/School/College
Knowledge Area
Date of Defense
Published
Ribeirão Preto, 2021
Supervisor
Committee
Barbosa, Marcello Henrique Nogueira (President)
Garcia, Luis Vicente
Guimarães, João Antonio Matheus
Kitamura, Felipe Campos
Title in Portuguese
Modelo radiômico baseado em imagens por tomografia computadorizada como preditor de metástase pulmonar em osteossarcoma
Keywords in Portuguese
Metástases
Osteosarcoma
Radiômica
Tomografia computadorizada
Abstract in Portuguese
O osteossarcoma (OS) é a neoplasia óssea primária maligna mais comum em crianças e adultos jovens, caracterizada por alta agressividade biológica e taxas elevadas de mortalidade, particularmente nos pacientes com doença metastática, que apresentam índices de sobrevida em cinco anos menor do que 60%. A radiômica é um processo complexo de análise quantitativa dos exames de imagem com potencial para inferir o comportamento biológico das lesões, servindo como uma ferramenta promissora, melhorando a acurácia diagnóstica, prognóstica e preditiva no acompanhamento dos pacientes. O objetivo do estudo foi analisar se um modelo radiômico de avaliação quantitativa de imagens obtidas por tomografia computadorizada pode ser preditor do surgimento de metástases pulmonares em pacientes com osteossarcoma. Trata-se de estudo retrospectivo no período de 2012 a 2021 que incluiu 150 pacientes com diagnóstico histopatológico confirmado de osteossarcoma. Foram incluídos os pacientes que realizaram tomografia computadorizada da lesão primária e do tórax antes de iniciar o tratamento quimioterápico e tomografia de tórax durante o seguimento a cada seis meses por dois anos. Diversos modelos de aprendizagem de máquina foram utilizados para predizer o surgimento de metástases pulmonares. A metástase pulmonar foi um importante fator relacionado à menor sobrevida dos pacientes com osteossarcoma. As variáveis LDH e gênero masculino foram identificadas como preditores de menor sobrevida e o tamanho da lesão mostrou associação com o desenvolvimento de metástase pulmonar. O desempenho de cada algoritmo foi avaliado pela área sob a curva (AUC), acurácia, sensibilidadade e especificidade. Entre outros resultados a pesquisa mostrou que o melhor modelo de classificação foi o Random Forest, com AUC de 0,79, acurácia de 73%, intervalo de confiança (IC) 95% [80-99%], sensibilidade de 71%, IC 95% [73%-100%] e especificidade de 83%, IC 95% [69-99%] na coorte de testagem final do modelo. Os atributos do modelo radiômico (assinatura radiômica) foram derivados de filtros wavelet e de transformada de Laplace da Gaussiana. Concluímos que o modelo radiômico por tomografia computadorizada utilizando o aprendizado de máquina pode auxiliar na avaliação do risco de desenvolvimento de metástases pulmonares nos pacientes com osteossarcoma.
Title in English
Radiomic model based on computed tomography images as a predictor of pulmonary metastasis in osteosarcoma
Keywords in English
Computed tomography
Metastases
Osteosarcoma
Radiomics
Abstract in English
Osteosarcoma (OS) is the most common primary malignant bone neoplasm in children and young adults, characterized by high biological aggressiveness and high mortality rates, particularly in patients with metastatic disease, who have five-year survival rates of less than 60%. Radiomics is a complex process of quantitative analysis of imaging exams with the potential to infer the biological behavior of lesions, serving as a promising tool, improving diagnostic, prognostic, and predictive accuracy in patient follow-up. The aim of the study was to analyze whether a radiographic model for quantitative assessment of images obtained by computed tomography can be a predictor of the appearance of pulmonary metastases in patients with osteosarcoma. This is a retrospective study from 2012 to 2021 that included 150 patients with a confirmed histopathological diagnosis of osteosarcoma. Patients who underwent computed tomography of the primary lesion and chest before starting chemotherapy treatment and chest tomography during follow-up every six months for two years were included. Several machine learning models were used to predict the onset of lung metastases. Pulmonary metastasis was an important factor related to the shorter survival of patients with osteosarcoma. The variables LDH and male gender were identified as predictors of shorter survival, and the size of the lesion was associated with the development of pulmonary metastasis. The performance of each algorithm was evaluated by the area under the curve (AUC), accuracy, sensitivity and specificity. Among other results, the research showed that the best classification model was the Random Forest, with AUC of 0.79, accuracy of 73%, confidence interval (CI) 95% [80-99%], sensitivity of 71%, CI 95% [73%-100%] and specificity 83%, 95% CI [69-99%] in the final model testing cohort. The attributes of the radiomic model (radiomic signature) were derived from wavelet filters and the Gaussian Laplace transform. We conclude that the computed tomography radiographic model using machine learning can help to assess the risk of developing lung metastases in patients with osteosarcoma.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2022-04-14
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.