Desarrollo de una herramienta de segmentación automática de los lóbulos pulmonares y del cálculo de la perfusión pulmonar en imágenes de TC y SPECT

Abstract

[ES] Las patologías pulmonares cada vez afectan a más población hasta el punto de que se estima que la Enfermedad Pulmonar Obstructiva Crónica (EPOC) se va a convertir en la tercera causa de muerte a nivel mundial en la siguiente década. Estas patologías afectan directa e indirectamente a la vascularización de los pulmones, por lo que los exámenes de perfusión pulmonar están cobrando importancia. La perfusión pulmonar se trata de una prueba de diagnóstico que permite evaluar la vascularización de los pulmones mediante imágenes SPECT. Junto a esta prueba, se acostumbra a realizar un escáner TC con la que se puede visualizar la anatomía pulmonar del paciente. Actualmente, se está llevando a cabo la cuantificación de la perfusión pulmonar mediante la división por tercios de la imagen SPECT y no por lóbulos, que es la morfología natural del pulmón. Por lo que los resultados obtenidos no son los más precisos en lo que respecta a la ubicación morfológica de la perfusión pulmonar. Por esta razón, en el presente trabajo se muestra una mejora de esta técnica mediante el desarrollo de una herramienta que permita la segmentación lobular morfológica de los pulmones, la fusión con la información obtenida con el SPECT pulmonar y posterior cuantificación de los valores de perfusión. Se ha desarrollado un breve estudio del estado del arte de las técnicas de segmentación y el posterior desarrollo de una herramienta que permita la segmentación lobular automatizada partiendo de imágenes TC. Para ello, se han utilizado las herramientas de segmentación lobular PTK y CIP. Posteriormente, se ha implementado la segmentación a la imagen SPECT y, por último, se han cuantificado los valores de perfusión pulmonar en la sección segmentada. Los resultados obtenidos se han comparado con los obtenidos en un estudio previo de la Unidad de Medicina Nuclear del Grupo ASCIRES donde la segmentación fue realizada manualmente por un especialista.

[EN] Lung diseases are increasingly affecting more people to the point that COPD is estimated to be the third leading cause of death worldwide in the next decade. These diseases directly and indirectly affect the vascularization of the lungs, so pulmonary perfusion tests are becoming increasingly important. Pulmonary perfusion is a diagnostic test that evaluates the vascularization of the lungs using SPECT imaging. Along with this test, a CT scan is usually performed to visualize the patient's lung anatomy. Currently, the quantification of lung perfusion is being carried out by dividing the SPECT image by thirds and not by lobes, which is the natural morphology of the lung. Therefore, the results obtained are not the most accurate regarding the morphological location of the lung perfusion. For this reason, the present work shows an improvement of this technique through the development of a tool that allows the morphological lobular segmentation of the lungs, fusion with the information obtained with the pulmonary SPECT and subsequent quantification of the perfusion values. A brief study of the state of the art of the segmentation techniques and the subsequent development of a tool that allows automated lobular segmentation from CT images has been developed. For this purpose, PTK and CIP lobe segmentation tools have been used. Subsequently, the segmentation has been implemented to the SPECT image and, finally, the pulmonary perfusion values in the segmented section have been quantified. The results obtained have been compared with those obtained in a previous study by the Nuclear Medicine Unit of the ASCIRES Group where the segmentation was performed manually by a specialist.

[CA] Lung diseases are increasingly affecting more people to the point that COPD is estimated to be the third leading cause of death worldwide in the next decade. These diseases directly and indirectly affect the vascularization of the lungs, so pulmonary perfusion tests are becoming increasingly important. Pulmonary perfusion is a diagnostic test that evaluates the vascularization of the lungs using SPECT imaging. Along with this test, a CT scan is usually performed to visualize the patient's lung anatomy. Currently, the quantification of lung perfusion is being carried out by dividing the SPECT image by thirds and not by lobes, which is the natural morphology of the lung. Therefore, the results obtained are not the most accurate regarding the morphological location of the lung perfusion. For this reason, the present work shows an improvement of this technique through the development of a tool that allows the morphological lobular segmentation of the lungs, fusion with the information obtained with the pulmonary SPECT and subsequent quantification of the perfusion values. A brief study of the state of the art of the segmentation techniques and the subsequent development of a tool that allows automated lobular segmentation from CT images has been developed. For this purpose, PTK and CIP lobe segmentation tools have been used. Subsequently, the segmentation has been implemented to the SPECT image and, finally, the pulmonary perfusion values in the segmented section have been quantified. The results obtained have been compared with those obtained in a previous study by the Nuclear Medicine Unit of the ASCIRES Group where the segmentation was performed manually by a specialist.