| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 6.57 MB | Adobe PDF |
Abstract(s)
Atualmente, a Cintigrafia Cerebral com DaTSCANTM SPECT é uma ferramenta que permite
o auxílio no diagnóstico de Síndromes Parkinsonianos. É um exame da Medicina Nuclear
(MN) cujas imagens são analisadas, na sua maioria, por inspeção visual de clínicos
experientes. Em casos duvidosos, de forma a auxiliar o diagnóstico, existem aplicações
comercialmente disponíveis que permitem a quantificação destes estudos a partir da
segmentação do tecido estriado cerebral, maioritariamente de forma semi-automática com
ajustes manuais por parte de um operador, cuja variabilidade intra e interoperador pode
influenciar a quantificação destes estudos e, em última análise, o diagnóstico clínico.
De forma a colmatar este problema, o objetivo deste trabalho foi construir uma aplicação
automática em linguagem de programação Python® para a análise de cortes transaxiais dos
corpos estriados em imagens obtidas por [123I]FP-CIT SPECT, provenientes do serviço de
MN do Hospital Particular de Almada (HPA), constituído por 68 pacientes, com e sem
patologia. De modo a comparar os rácios de captação específica obtidos para os corpos
estriados, desenvolveu-se ainda uma aplicação semi-automática com ajustes manuais por
parte de um operador. Finalmente os resultados obtidos foram comparados com a aplicação
DaTSCAN V4, da General Electric Healthcare. Verificou-se que a aplicação automática
desenvolvida consegue discriminar com sucesso os pacientes sem patologia dos pacientes
com patologia nos rácios A (Corpos estriados/Occipital) (AUC=0,85, sensibilidade=80%,
especificidade=90%), B (Corpo estriado esquerdo/Occipital) (AUC=0,82, sensibilidade=84%,
especificidade=80%) e C (Corpo estriado direito/Occipital) (AUC=0,85, sensibilidade=80%,
especificidade=90%), com forte correlação com os métodos semi-automáticos (DaTSCANV4
e aplicação desenvolvida). Contudo, esta correlação não se observou para o rácio D (Corpo
estriado esquerdo/Corpo estriado direito) (AUC=0,63, sensibilidade=56%,
especificidade=70%). Verificou-se também a existência de variabilidade intra e interoperador
nos métodos semi-automáticos, contrariamente à aplicação automática desenvolvida.
A implementação deste método para segmentação e quantificação de cortes transaxiais dos
corpos estriados em imagens obtidas por [123I]FP-CIT SPECT demonstrou resultados
tendencialmente promissores, cuja potencialidade de aplicação pode complementar
diferencialmente a análise visual, ainda que careça de otimização e validação futura.
Currently, a Brain SPECT with DaTSCANTM is a tool that helps diagnose of Parkinsonian Syndromes. It is an imaging technique of Nuclear Medicine, whose images are mostly analized by visual interpretation from experienced observers. In doubtful cases, in order to help the diagnosis, there are commercial software packages that allow the quantification of these studies from the segmentation of the striatum, mostly by semi-quantitative measures with manual adjustments from an operator, whose intra and inter-operator variability can influence the quantification of these studies and clinical diagnosis. In order to overcome this problem, the main goal of this project is to construct an automatic application in Python for the analysis of transaxial slices of the striatum in images obtained by [123I] FP-CIT SPECT from nuclear medicine department of the Hospital Particular de Almada, with 68 patients with and without pathology. In order to compare the specific uptake ratios for the striatum, a semi-automatic application was developed with manual adjustments by an operator. Finally the obtained results were compared with the DaTSCAN V4 application, from General Electric Healthcare. It was possible to verify that the automatic application developed successfully discriminates the healthy patients of the pathological ones in the ratios A (Striatum/Occipital) (AUC=0.85, sensitivity=80%, specificity=90%), ratio B (Left Striatum/Occipital) (AUC=0.82, sensitivity=84%, specificity=80%), and ratio C (Right Striatum/Occipital) (AUC=0.85, sensitivity=80%, specificity=90%), with a strong correlation with semi-automatic methods (DaTSCAN V4 and developed application). However, this correlation was not observed for the ratio D (Left Striatum/Right Striatum) (AUC=0.63, sensitivity=56%, specificity=70%). There was also an existence of intra- and inter-operator variability in semi-automatic methods, contrary to the automatic application developed. The implementation of this method for segmentation and quantification of transaxial slices of the striatum in images obtained by [123I]FP-CIT SPECT demonstrated promising results, whose potentiality of application can complement the visual analysis differently, even though it lacks future optimization and validation.
Currently, a Brain SPECT with DaTSCANTM is a tool that helps diagnose of Parkinsonian Syndromes. It is an imaging technique of Nuclear Medicine, whose images are mostly analized by visual interpretation from experienced observers. In doubtful cases, in order to help the diagnosis, there are commercial software packages that allow the quantification of these studies from the segmentation of the striatum, mostly by semi-quantitative measures with manual adjustments from an operator, whose intra and inter-operator variability can influence the quantification of these studies and clinical diagnosis. In order to overcome this problem, the main goal of this project is to construct an automatic application in Python for the analysis of transaxial slices of the striatum in images obtained by [123I] FP-CIT SPECT from nuclear medicine department of the Hospital Particular de Almada, with 68 patients with and without pathology. In order to compare the specific uptake ratios for the striatum, a semi-automatic application was developed with manual adjustments by an operator. Finally the obtained results were compared with the DaTSCAN V4 application, from General Electric Healthcare. It was possible to verify that the automatic application developed successfully discriminates the healthy patients of the pathological ones in the ratios A (Striatum/Occipital) (AUC=0.85, sensitivity=80%, specificity=90%), ratio B (Left Striatum/Occipital) (AUC=0.82, sensitivity=84%, specificity=80%), and ratio C (Right Striatum/Occipital) (AUC=0.85, sensitivity=80%, specificity=90%), with a strong correlation with semi-automatic methods (DaTSCAN V4 and developed application). However, this correlation was not observed for the ratio D (Left Striatum/Right Striatum) (AUC=0.63, sensitivity=56%, specificity=70%). There was also an existence of intra- and inter-operator variability in semi-automatic methods, contrary to the automatic application developed. The implementation of this method for segmentation and quantification of transaxial slices of the striatum in images obtained by [123I]FP-CIT SPECT demonstrated promising results, whose potentiality of application can complement the visual analysis differently, even though it lacks future optimization and validation.
Description
Trabalho final de mestrado para obtenção do grau de mestre em Engenharia Biomédica
Keywords
Segmentação de imagem Image segmentation Quantificação Quantification Rácios de captação específica Specific uptake ratios Análise de imagem automática Automated image analysis DaTSCANTM
Citation
ELIAS, Maria Madalena Penaforte Pirão - Aplicação para análise de cortes transaxiais dos corpos estriados em imagens obtidas por [123|]FP-CIT SPECT. 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