Evaluación de la calidad de imagen del PET/CT Biograph Vision 450 al variar parámetros de reconstrucción
Fecha
2023
Tipo
tesis de maestría
Autores
Montero Alpízar, Sebastián
Título de la revista
ISSN de la revista
Título del volumen
Editor
Resumen
Los datos obtenidos por los sistemas PET/CT pasan por un algoritmo de reconstrucción antes de poder generar las imágenes. Estos algoritmos de reconstrucción permiten variar parámetros como el número de iteraciones y filtrado post reconstrucción. El uso de diferentes combinaciones de dichos parámetros resulta en imágenes con distintos ruidos, contrastes y resolución espacial. Este trabajo ofrece una evaluación del efecto de estos parámetros en la calidad de la imagen para lesiones hipocaptantes e hipercaptantes simuladas por un maniquí para el PET/CT Biograph Vision 450.
Se evalúa el método de reconstrucción UltraHD PET en combinación con filtros Gaussianos de 2.5mm, 5mm, 10mm 15mm y la opción All-Pass que no aplica ningún filtro, todas estas con 1, 5, 10, 15, 20, 30, 40, 50 iteraciones, manteniendo el número de subgrupos en 5. Se utiliza un maniquí de control de calidad SPECT Jaszczak con esferas solidas e inserto de varillas para simular las lesiones hipocaptantes y un maniquí de cuerpo EIC NEMA PET con esferas rellenables que se llenaron con flúor 18 para simular lesiones hipercaptantes. A ambos maniquíes se les agrega una región con actividad de fondo. Se evalúa la razón contraste ruido, coeficiente de recuperación de contraste, variabilidad de fondo, rugosidad de la imagen y resolución espacial.
Se determina que el CRC, BV y IR incrementan con el número de iteraciones y que disminuyen al incrementar el tamaño (FWHM) del filtro Gaussiano. La CNR disminuye al aumentar el número de iteraciones, pero mantiene un valor constante con los filtros de 10 y 15mm. Al relacionar el aumento del ruido con el HCRC se obtuvo que los mejores resultados se obtienen al utilizar el filtro de 2.5mm con 15 iteraciones. Se obtiene que la resolución espacial mejora al incrementar el número de iteraciones, para la lesión hipocaptante se encuentra que se empeora la resolución espacial al incrementar el tamaño de filtro. Para la lesión hipercaptante se obtienen valores del FWHM más cercanos al tamaño de la esfera para el filtro de 10mm.
The data obtained by PET/CT systems goes through a reconstruction algorithm before generating the images. These reconstruction algorithms allow for varying parameters such as the number of iterations and post-reconstruction filtering. The use of different combinations of these parameters results in images with different noise, contrast, and spatial resolution. This work offers an evaluation of the effect of these parameters on image quality for hot and cold lesions simulated by a mannequin for the PET/CT Biograph Vision 450. The UltraHD PET reconstruction method is evaluated in combination with Gaussian filters of 2.5mm, 5mm, 10mm, 15mm, and the All-Pass option that applies no filter, all with 1, 5, 10, 15, 20, 30, 40, 50 iterations, while maintaining the number of subgroups at 5. A SPECT Jaszczak quality control phantom with solid spheres and rod inserts is used to simulate cold lesions, and an EIC NEMA PET body phantom with fillable spheres filled with fluorine-18 is used to simulate hot lesions. Both phantoms have a background activity region added. The contrast-to-noise ratio, contrast recovery coefficient, background variability, image roughness, and spatial resolution are evaluated. It is determined that CRC, BV, and IR increase with the number of iterations and decrease as the size (FWHM) of the Gaussian filter increases. The CNR decreases as the number of iterations increases but maintains a constant value with 10mm and 15mm filters. By relating the increase in noise to the HCRC, it was found that the best results are obtained when using the 2.5mm filter with 15 iterations. It is noted that the spatial resolution improves with an increase in the number of iterations. In the case of the hot lesion, the spatial resolution is found to worsen as the filter size increases. Conversely, for the cold lesion, FWHM values closer to the sphere size are obtained with the 10mm filter.
The data obtained by PET/CT systems goes through a reconstruction algorithm before generating the images. These reconstruction algorithms allow for varying parameters such as the number of iterations and post-reconstruction filtering. The use of different combinations of these parameters results in images with different noise, contrast, and spatial resolution. This work offers an evaluation of the effect of these parameters on image quality for hot and cold lesions simulated by a mannequin for the PET/CT Biograph Vision 450. The UltraHD PET reconstruction method is evaluated in combination with Gaussian filters of 2.5mm, 5mm, 10mm, 15mm, and the All-Pass option that applies no filter, all with 1, 5, 10, 15, 20, 30, 40, 50 iterations, while maintaining the number of subgroups at 5. A SPECT Jaszczak quality control phantom with solid spheres and rod inserts is used to simulate cold lesions, and an EIC NEMA PET body phantom with fillable spheres filled with fluorine-18 is used to simulate hot lesions. Both phantoms have a background activity region added. The contrast-to-noise ratio, contrast recovery coefficient, background variability, image roughness, and spatial resolution are evaluated. It is determined that CRC, BV, and IR increase with the number of iterations and decrease as the size (FWHM) of the Gaussian filter increases. The CNR decreases as the number of iterations increases but maintains a constant value with 10mm and 15mm filters. By relating the increase in noise to the HCRC, it was found that the best results are obtained when using the 2.5mm filter with 15 iterations. It is noted that the spatial resolution improves with an increase in the number of iterations. In the case of the hot lesion, the spatial resolution is found to worsen as the filter size increases. Conversely, for the cold lesion, FWHM values closer to the sphere size are obtained with the 10mm filter.
Descripción
Palabras clave
EVALUACIÓN, CALIDAD DE IMAGEN, METODO DE RECONSTRUCCIÓN, PARÁMETRO