2006 Fiscal Year Final Research Report Summary
Recognition of fine structure by the spatiotemporal tracking for multi-dimensional images
| Project/Area Number |
16560366
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Measurement engineering
|
|---|
| Research Institution | Chukyo University |
|---|
Principal Investigator |
MEKADA Yoshito Chukyo University, School of Life system Science and Technology, Professor, 生命システム工学部, 教授 (00282377)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
TORIWAKI Junichiro Chukyo University, School of Life system Science and Technology, Professor, 生命システム工学部, 教授 (30023138)
HASEGAWA Jun-ichi Chukyo University, School of Life system Science and Technology, Professor, 生命システム工学部, 教授 (30126891)
MURASE Hiroshi Nagoya univ., Graduate School of Information Science, Professor, 大学院情報科学研究科, 教授 (90362293)
MORI Kensaku Nagoya univ., Graduate School of Information Science, Associate Professor, 大学院情報科学研究科, 助教授 (10293664)
|
|---|
| Project Period (FY) |
2004 – 2006
|
| Keywords | Pattern recognition / Multi-dimensional images / Image features / Medical X-ray images / Segmentation / Registration |
|---|
| Research Abstract |
Recent progress in such CT imaging devices as multidetector-row CT scanners enables us to take more precise slice images of a patient. But this huge number of slice images also increases the burden on doctors to diagnose. The development of a computer-aided diagnosis (CAD) system has reduced such burdens on doctors and the quantification of diagnosing. CAD systems for the lungs must include the following functions: (a) detection of such suspicious regions as tumors, and (b) the discrimination of benignancy or malignancy of those suspicious regions. When discriminating between benign or malignant lung tumors, the kind of pulmonary blood vessels involved in tumors is very important. For adenocarcinoma, since tumors tend to be located around the interlobar and pulmonary veins positioned roughly in the center of tumors, the kind of vessel is useful supplemental information to discriminate lung tumors with computers. To get such information, we have developed a method for the automatic recognition of pulmonary arteries and veins by using anatomical positional relationships between each bronchus and vessel. As the results, in the best cases, about 95% of the vessels were classified correctly by the proposed method. Even in the worst cases, more than 80% of the blood vessels were classified correctly.
In the diagnosis of a liver, a doctor examines dynamic CT images. These consist of four images, namely the pre-contrast phase, early phase, portal phase, and late phase ones, which are taken sequentially within a few minutes. Since the early and late phase images are important for diagnosing liver cancer, we propose a method to analyze spatiotemporal domain to such precise image. Our method refers to both of them for detecting suspicious regions and eliminating false positives. Experimental results showed that the number of false positive regions per case was 0.33 with sensitivity of 100%.
|
