| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 2.18 MB | Adobe PDF |
Authors
Abstract(s)
Os doentes com suspeita de trombo-embolismo pulmonar são frequentemente avaliados
por técnicas cintigráficas, cujas limitações têm estimulado necessidade de criar aplicações de
processamento de imagem para melhorar o suporte de análise no diagnóstico clínico.
Pretendeu-se construir um método de processamento de imagem para extrair características
das imagens da cintigrafia de ventilação e perfusão pulmonar (CVPP), partindo de um estudo
retrospetivo com 71 indivíduos (média de idades 66,1 ±13,8 anos) que realizaram CVPP no
Hospital Particular de Almada por indicação clínica (09/2014 a 06/ 2017). Aplicou-se aos
estudos o método de processamento de imagem para CVPP desenvolvido com base num
algoritmo automático e semiautomático (permite ajustes por parte do operador) para extração
de características. Na análise estatística utilizou-se testes não paramétricos para comparar
variáveis (significância 5%). Comparou-se as variáveis das amostras de doentes saudáveis e
com patologia pulmonar através de curvas ROC. Incluiu-se a área do pulmão na região de
interesse nos dois métodos de processamento. A análise estatística revela um maior número
de casos onde os valores médios das variáveis em estudo obtêm melhores resultados quando
o processamento é feito de modo semiautomático e nos estudos de perfusão. Distinguiu-se
as probabilidades de existência de patologia identificadas em relatório médico na amostra
patológica de forma global e localizou-se as mesmas nos diferentes lóbulos pulmonares. A
metodologia desenvolvida permitiu incluir no processo de segmentação toda a região de
interesse independentemente do método de processamento. Apesar da necessidade de
realizar ajustes por parte do operador, este método mostrou-se promissor na avaliação destas
populações. Carece, no entanto, de uma amostra maior e uma análise futura da variabilidade
intraoperador, para se afirmar com propriedade os resultados.
Patients with suspected pulmonary thromboembolism are often assessed by scintigraphic techniques, which may have some restrictions that encourage the need to create image processing applications to improve the support of analysis in clinical diagnosis. The intention is to construct an image processing method to extract characteristics of pulmonary ventilation and perfusion scintigraphy (PVPS), starting from a retrospective study conducted with 71 subjects (mean age 66.1 ±13,8 years) who underwent PVPS in the Private Hospital of Almada by clinical indication (09/2014 to 06/2017). The image processing method for PVPS was developed based on an automatic and semi-automatic algorithm (which allows adjustments by the operator) for extraction of characteristics. In the statistical analysis, non-parametric tests were used to compare variables (5% significance). The variables of the pathological and healthy samples were compared using ROC curves. The lung area was icluded in the region of interest in both processing methods. Statistical analysis reveals a greater number of cases where the mean values of the variables in study obtain better results when the processing is semiautomatic and in the perfusion studies. The probability of existence of pathology identified in a medical report in the pathological sample was globally distinguished and the same were found in the different pulmonary lobes. The results allowed to include the entire region of interest in the segmentation process regardless of the processing method. Despite the need for adjustments by the operator, this method proved itself promising in the assessment of these populations. It lacks, however, a larger sample and a future analysis of the variability intraoperator, to state the results.
Patients with suspected pulmonary thromboembolism are often assessed by scintigraphic techniques, which may have some restrictions that encourage the need to create image processing applications to improve the support of analysis in clinical diagnosis. The intention is to construct an image processing method to extract characteristics of pulmonary ventilation and perfusion scintigraphy (PVPS), starting from a retrospective study conducted with 71 subjects (mean age 66.1 ±13,8 years) who underwent PVPS in the Private Hospital of Almada by clinical indication (09/2014 to 06/2017). The image processing method for PVPS was developed based on an automatic and semi-automatic algorithm (which allows adjustments by the operator) for extraction of characteristics. In the statistical analysis, non-parametric tests were used to compare variables (5% significance). The variables of the pathological and healthy samples were compared using ROC curves. The lung area was icluded in the region of interest in both processing methods. Statistical analysis reveals a greater number of cases where the mean values of the variables in study obtain better results when the processing is semiautomatic and in the perfusion studies. The probability of existence of pathology identified in a medical report in the pathological sample was globally distinguished and the same were found in the different pulmonary lobes. The results allowed to include the entire region of interest in the segmentation process regardless of the processing method. Despite the need for adjustments by the operator, this method proved itself promising in the assessment of these populations. It lacks, however, a larger sample and a future analysis of the variability intraoperator, to state the results.
Description
Trabalho final de mestrado para obtenção do grau de mestre em Engenharia Biomédica
Keywords
Processamento de imagem Image processing Quantificação Quantification Segmentação Segmentation Cintigrafia de ventilação perfusão pulmonar Pulmonary ventilation perfusion scintigraphy Imagem médica Medical imaging
Citation
LINARES, Inês Micaela de Jesus - Construção de método de processamento para análise automática e semiautomática de imagens de cintigrafia de ventilação perfusão pulmonar. Lisboa: Instituto Superior de Engenharia de Lisboa - Escola Superior de Tecnologia da Saúde de Lisboa, 2017. Dissertação de mestrado.
Publisher
Instituto Superior de Engenharia de Lisboa - Escola Superior de Tecnologia da Saúde de Lisboa