Texture multichannel measurements for cancer precursors’ identification using support vector machines
| Autor: | Dimitris Maroulis, S.A. Karkanis, Dimitris K. Iakovidis, Panagiotis Papageorgas, M. Tzivras |
|---|---|
| Rok vydání: | 2004 |
| Předmět: |
Computer science
Colorectal cancer business.industry Applied Mathematics Frame (networking) Cancer Wavelet transform Covariance Condensed Matter Physics medicine.disease Support vector machine Color model Wavelet medicine Computer vision Artificial intelligence Electrical and Electronic Engineering business Instrumentation |
| Zdroj: | Measurement. 36:297-313 |
| ISSN: | 0263-2241 |
| DOI: | 10.1016/j.measurement.2004.09.010 |
| Popis: | Colorectal cancer is one of the leading types of cancer in the developed countries. Epidemiological studies have shown that the risk of developing colorectal cancer can be significantly reduced through early detection and removal of cancer precursor lesions. We propose a novel framework for the automated identification of colon cancer precursors based on the processing of color video frames acquired during endoscopy. The spectral information of the three color channels forming the endoscopic frames is used for the description of the colonic mucosa. The suitability of different color models for this application is investigated. The textural properties of the colonic mucosal surface are measured using second order statistical descriptors on the wavelet transform of the multichannel video signals. A new reduced set of measures based on the inter-channel covariance of the features has proven to provide high discrimination of image regions corresponding to normal and abnormal tissue. The proposed framework was tested using a Support Vector Machine classifier on different video frame sets presenting adenomatous polyps and the average sensitivity and specificity was estimated to reach 94% and 95.7%, respectively. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect