2004 Fiscal Year Final Research Report Summary
A basic research on development of a CAD system for multi-diseases in 3-D CT images
| Project/Area Number |
15591263
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Radiation science
|
|---|
| Research Institution | Gifu University |
|---|
Principal Investigator |
FUJITA Hiroshi Gifu University, Graduate School of Medicine, Professor, 大学院・医学研究科, 教授 (10124033)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
HOSHI Hiroaki Gifu University, Graduate School of Medicine, Professor, 大学院・医学研究科, 教授 (60128395)
HARA Takeshi Gifu University, Graduate School of Medicine, Associate Professor, 大学院・医学研究科, 助教授 (10283285)
ZHOU Xiangrong Gifu University, Graduate School of Medicine, Research Associate, 大学院・医学研究科, 助手 (00359738)
NISHIHARA Sadamitsu Hiroshima Prefectural College of Health Sciences, Research Associate, 保健福祉学部, 助手 (40290548)
ISHIGAKI Takeo Nagoya University, Graduate School of Medicine, Professor, 大学院・医学研究科, 教授 (60094356)
|
|---|
| Project Period (FY) |
2003 – 2004
|
| Keywords | Computer-aided diagnosis / Medical image processing / Medical image recognition / CT / Computer-aided detection / International information exchange (USA) |
|---|
| Research Abstract |
The aim of this study is to perform a basic research on developing a computer-aided diagnosis (CAD) system for multi-diseases, especially existing in thoracic and abdominal regions, in 3-dimentional (3D) multi-slice CT images and to investigate its applicability to clinical situations. We have obtained the results as shown below.
1)Collection of 3D X-ray CT images (construction of CT image database) : About 200 cases of 3D CT image data of torso regions as well as brain vessel CT Images are collected and stored in RAID discs in the laboratory. All of these image data consist of normal parts of the human body (originally imaged because of existence of suspected disease(s)), pulmonary cancer, emphysema, or other abnormalities such as liver disease.
2)Development of a scheme for automatically recognizing anatomical normal structure : An algorithm for segmenting main internal organs fully automatically in torso regions was developed and evaluated the accuracy and precision in terms of segmentation and recognition by using the database described above. As an overall result, satisfied result was obtained with segmentation accuracy of larger than 70 %.
3)Development of a scheme for automatically detecting abnormal regions : Thoracic lesion (pulmonary nodule), emphysema, breast cancer (mass shadow), liver lesion (cirrhosis), colon polyp, and osteoporosis were automatically detected by the developed algorithms and we found that these initial results were accepted with detection accuracy of larger than 80 %).
4)We obtained successful results as shown in 2) and 3) for initial CAD schemes for detecting and recognizing mufti-diseases from 3D torso multi-slice CT images, and this will be merged as a CAD system for multi-diseases ; we are still working on this part.
|
