Home // PATTERNS 2017, The Ninth International Conferences on Pervasive Patterns and Applications // View article

Polyp Classification Using Multiple CNN-SVM Classifiers from Endoscope Images

Authors:
Masataka Murata
Hiroyasu Usami
Yuji Iwahori
Wang Aili
Naotaka Ogasawara
Kunio Kasugai

Keywords: Polyp Classification; Endoscope Image; Voting Processing; Pre-Trained Network; Convolutional Neural Network; Support Vector Machine.

Abstract:
This paper proposes a classification approach of a malignant or bening polyp type by multiple CNN-SVM classifiers using Convolutional Neural Networks (CNN) as feature extractor and Support Vector Machine (SVM) as classifier from three kinds of endoscope images as white light image, dye image and Narrow Band Image (NBI). First, the polyp feature is extracted using CNN as feature extractor from three kinds of endoscope images using each datasets. Second, classifiers are generated as many as three kinds of combinations using SVM and each image is classified. Finally, the final classification result is judged by voting processing from the result obtained by each classifier. The effectiveness of the proposed method was confirmed through experiments in which both validity and accuracy of multiple CNN-SVM voting results are evaluated using actual malignant or benign polyp images.

Pages: 109 to 112

Copyright: Copyright (c) IARIA, 2017

Publication date: February 19, 2017

Published in: conference

ISSN: 2308-3557

ISBN: 978-1-61208-534-0

Location: Athens, Greece

Dates: from February 19, 2017 to February 23, 2017