Estimation of gastrointestinal polyp size in video endoscopy

Abstract Worldwide the colorectal cancer is one of the most common public health problems, constituting in 2010 the seventh cause of death. This aggressive cancer is firstly identified during an endoscopy routine examination by characterizing a set of polyps that appear along the digestive tract, mainly in the colon and rectum. The polyp size is one of the most important features that determines the surgical endoscopy management and even can be used to predict the level of aggressiveness. For instance, the gastroenterologists only send a polyp sample to the pathology examination if the polyp diameter is larger than 10 mm, a measure that is achieved typically by examining the lesion with a calibrated endoscopy tool. However, the polyp size measure is very challenging because it must be performed during a procedure subjected to a complex mix of noise sources, such as: the distorted optical characteristics of the endoscopy, the exacerbated physiological conditions and abrupt motion. The main goal of this thesis was estimated the polyp size during an endoscopy video sequence using a spatio-temporal characterization. Firstly, the method estimated the region with more motion within which the polyp shape is approximated by those pixels with the largest temporal variance. On the above, an initial manual polyp delineation in the first frame captures the main features to be follow in posterior frames by a cross correlation procedure. Afterwards, a bayesian tracking strategy is used to refine the polyp segmentation. Finally a defocus strategy allows to estimate on the clear cut frame at a certain depth as a reference to determine the polyp size obtaining reliable results. In the segmentation task, the approach achieved a Dice Score of 0.7 in real endoscopy video-sequences, when comparing with an expert. In addition, the results polyp size estimation obtained a Root Mean Square Error (RMSE) of 0.87 mm with spheres of known size that simulated the polyps, and in real endoscopy sequences obtaining a RMSE of 4.7 mm compared with measures obtained by a group of four experts with similar experience.

• Autores: Ruano Balseca, Josué André
• Fecha de publicación: 2015
• Institución: Universidad Nacional de Colombia
• Repositorio: Universidad Nacional de Colombia
• Idioma: spa
• OAI Identifier: oai:repositorio.unal.edu.co:unal/54703
• Acceso en línea: https://repositorio.unal.edu.co/handle/unal/54703; http://bdigital.unal.edu.co/49820/
• Palabra clave: 5 Ciencias naturales y matemáticas / Science; 51 Matemáticas / Mathematics; 6 Tecnología (ciencias aplicadas) / Technology; 61 Ciencias médicas; Medicina / Medicine and health; Polyp size estimation; Endoscopy; Colorectal cancer; Polyp shape segmentation; Spatio-temporal characterization; Estimación del tamaño de los pólipos,; Endoscopía; Cancer colorectal; Segmentación de la forma de los pólipos; Caracterización espaciotemporal
• Rights: openAccess
• License: Atribución-NoComercial 4.0 Internacional