Automatic Polyp Recognition and Shape and Size Recovery for Medical Endoscopy Diagnosis

2016 Fiscal Year Final Research Report

• Project/Area Number: 26330210
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Perceptual information processing
• Research Institution: Chubu University
• Principal Investigator: IWAHORI Yuji 中部大学, 工学部, 教授 (60203402)
• Co-Investigator(Renkei-kenkyūsha): FUKUI Shinji 愛知教育大学, 教育学部, 准教授 (80345941) ; KASUGAI Kunio 愛知医科大学, 医学部, 教授 (80298570)
• Research Collaborator: Woodham Robert J. UBC, Professor ; Bhuyan M. K. IIT Guwahati, Associate Professor
• Project Period (FY): 2014-04-01 – 2017-03-31
• Keywords: コンピュータビジョン / 医療画像応用 / 知覚情報処理 / 内視鏡画像 / 3次元形状復元 / Shape from Shading / ポリープ検出 / ニューラルネット

Outline of Final Research Achievements

This research proposed approaches to recover the polyp shape and size from endoscope image. Using two images with known movement along Z direction can estimate the reflectance parameters of uniform diffused reflectance shading image which is converted from the original image, and can recover the absolute size of polyp.
Generating Lambertian image is also improved and observation model are considered and evaluated for one point light source and two point light source mdoel for shape recovery of polyp and it is shown that two point light source model gives better shape with estimating the absolute polyp size.
Polyp detection approach is also developed using edge information and color information, which makes likelihood map of polyp candidate region. HOG feature is taken from the polyp candidate region and machine learning approach is used to judge if the polyp candidate region is ture or false. Approach for classification of polyp with benign or malignant is also proposed.

Free Research Field

コンピュータビジョン