SYSTEM AND METHOD FOR DETECTING POLYPS FROM LEARNED BOUNDARIES

摘要

A system and method for automated polyp detection in optical colonoscopy images is provided. In one embodiment, the system and method for polyp detection is based on an observation that image appearance around polyp boundaries differs from that of other boundaries in colonoscopy images. To reduce vulnerability against misleading objects, the image processing method localizes polyps by detecting polyp boundaries, while filtering out irrelevant boundaries, with a generative-discriminative model. To filter out irrelevant boundaries, a boundary removal mechanism is provided that captures changes in image appearance across polyp boundaries. Thus, in this embodiment the boundary removal mechanism is minimally affected by texture visibility limitations. In addition, a vote accumulation scheme is applied that enables polyp localization from fragmented edge segmentation maps without identification of whole polyp boundaries.

主权项

1. A system for automated polyp detection in optical colonoscopy images, the system comprising:
an input configured to acquire a series of optical images; a processor in communication with the input; and a memory containing instructions that, when executed by the processor, causes the processor to perform a process on the series of optical images comprising:
applying a color filter to create a plurality of color filtered images for each optical image of the series of optical images;locating a series of edge pixels on at least one of the plurality of color filtered images;obtaining a plurality of oriented image patches corresponding to each of the series of the edge pixels, the plurality of oriented image patches represented by an intensity signal characterized by at least one of rotation invariance or illumination invariance;estimating an edge normal for each edge pixel;constructing at least one classification system corresponding to the plurality of color filtered images, the at least one classification system configured to enhance low level features of the plurality of color filtered images prior to classification;generating appearance features from a series of feature vectors selected from the at least one classification system;performing an edge classification threshold analysis on each of the plurality of oriented image patches;determining, based on the edge classification threshold analysis, that a given image patch is consistent with an edge threshold; andgenerating a report indicating potential polyps with a vote accumulation greater than a threshold, the vote accumulation including a probabilistic output for each of the potential polyps in the absence of a predefined parametric model.