Deep Learningを用いた内視鏡画像の分類と血管を利用したポリープ形状復元

機関リポジトリ

書誌事項

タイトル別名
  • Classification of Endoscope Image using Deep Learning and Shape Recovery of Polyp Based on Blood Vessel

説明

type:論文

By realizing the 3D-shape recovery of the image inside of the human body taken by the endoscope, the quantitative evaluation of polyps becomes possible. Polyp detection, classification, and shape recovery methodologies are required based on endoscope images taken by "general" endoscope images. Against "general" endoscope images, this research proposes an efficient polyp classification method using a convolutional neural network trained in advance for other tasks as a feature extractor and construct multiple SVM (Support Vector Machine) as a classifier. Furthermore, for polyps which are difficult to classify benign or malignant, a robust polyp shape recovery method is proposed based on blood vessel information.

収録刊行物

キーワード

詳細情報 詳細情報について

問題の指摘

ページトップへ