1995 Fiscal Year Final Research Report Summary
Computer-Aided Diagnosis System in Chest Radiography
| Project/Area Number |
06454326
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Radiation science
|
|---|
| Research Institution | Iwate Medical University |
|---|
Principal Investigator |
KATSURAGAWA Shigehiko Iwate Med.Univ., Radiology assistant prof., 医学部, 講師 (60021630)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
NAKASATO Tatsuhiko Iwate Med.Univ., Radiology research associate, 医学部, 助手 (10207748)
YOSHIOKA Kunihiro Iwate Med.Univ., Radiology assistant prof., 医学部, 講師 (70210648)
|
|---|
| Project Period (FY) |
1994 – 1995
|
| Keywords | Computer-Aided Diagnosis / Digital Radiography / Image Database / Chest Radiograph / Digital Image Processing |
|---|
| Research Abstract |
We have developed a system and schemes of computer-aided diagnosis (CAD) for detection of nodules, interstitial diseases, pneumothorax and cardiomegaly in digital chest radiograph.
1. The CAD system hardware was established, which includes a film-digitizer, a workstation, a laser-imager, a video-printer and a personal computer.
2. The image-difference method was developed for detection of lung nodules. The performance with approximately 75% of the sensitivity at 1.0 false positive/image was provided for 100 normal and 100 abnormal lungs with lung nodules.
3. The texture and geometric feature analyzes were developed for detection of interstitial diseases. The performance with approximately 95% of the sensitivity at 95% of the specificity was provided for 100 normal and 100 abnormal lungs with interstitial infiltrates.
4. The method with Hough transform was developed for detection of pneumothorax. 80% sensitivity at 70% specificity was provided for 60 lungs with pneumothorax.
5. The method based on a model function was developed for measurement of size and shape of hearts.
6. The ROC analysis showed that the diagnostic accuracy with both chest images and CAD outputs was superior to that with chest images alone for detection of nodules and interstitial diseases.
7. Our results indicates that this computer-aided diagnosis system is very useful for radiologists to estimate chest radiographs.
|
