| Project/Area Number |
03454281
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Radiation science
|
|---|
| Research Institution | Nagoya University |
|---|
Principal Investigator |
ENDO Tokiko NAGOYA UNIVERSITY, SCHOOL OF MEDICINE DEPT.OF RADIOLOGY ASSOCIATE PROFESSOR, 医学部, 助教授 (10231193)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
IKEDA Mituru NAGOYA UNIVERSITY, SCHOOL OF MEDICINE DEPT.OF MEDICAL INFORMATION ASSOCIATE PROF, 医学部, 助教授 (50184437)
FUJITA Hirosi GIFU UNIVERSITY, SCHOOL OF TECHNOLOGY DEPT.OF COMPUTER ENGINEERING ASSOCIATE PRO, 工学部・電子情報工学科, 助教授 (10124033)
佐久間 貞行 名古屋大学, 医学部, 教授 (90079963)
|
|---|
| Project Period (FY) |
1991 – 1993
|
| Project Status |
Completed (Fiscal Year 1993)
|
| Budget Amount *help |
¥6,000,000 (Direct Cost: ¥6,000,000)
Fiscal Year 1993: ¥500,000 (Direct Cost: ¥500,000)
Fiscal Year 1992: ¥500,000 (Direct Cost: ¥500,000)
Fiscal Year 1991: ¥5,000,000 (Direct Cost: ¥5,000,000)
|
|---|
| Keywords | mammography / breast cancer / auto-analyzing system / mass density / calcifications / digitized mammograms / neural network / cluster calcifications / マンモグラム / 自動診断 / 診断支援装置 / デジタル画像 / 石灰化像 / 腫瘤抽出 / 腫瘤の良悪性の鑑別 / ブレストファントム |
|---|
| Research Abstract |
We had developed an assistant diagnostic system for breast cancer in which the landmark was the mass density on digital mammography. And we have been working on the investigation for the classification of possible extracted tumors into benign and malignant ones using the artificial neural network technique. A sequential-dependence technique, which calculates the degree of redundancy or patterning in a sequence, was employed to extract image features from mammographic iamages. The extracted vectors were then used as input to the neural network. The results show that the neural network could correctly classify benign and malignant tumors with an average success rate of 85% for 40 mammograms (20 benign, 20 malignant). This accuracy rate indicated that the neural network approch had potentialy utility in the computer-aided diagnosis of breast cancer.
For detection of the microcalcifications, we have geen making two approaches. One is the determination of the spatial resolution required for
… More
providing enough detectability of mammographic microcalcifications. Radiographs of a breast phantom, contained simulated microcalcifications, were digitized by five pixel sizes from 25-500mum with 12-bits gray levels by a drum scanneer. then the images were evaluated by phisical image quality index, calculated from displayd amplitude model in detection process, and were also assessed by the visual image quality rank in a human observer performance study. The results were that a spatial resolution smaller than 100mum pixel size showed high or enough detectability of subtle microcalcifications on mammograms.
The other study is the investigation for detection the microcalcifications in the clinical screen/film mammograms. Using the laser scanneer at a pixel size of 100mum and 10-bit gray levels, we investigated a new scheme for the detection of microcalcifications in finely-sampled mammograms. The region extraction of the breast region, enhancement on the high frequency components, and the detection of microcalcifications and their cluster were done by computer system. The detection accuracy evaluated for sampled 39 cases was 87.2% and for clinical 163 cases was 74.2%.
We are planning on developing and improving of the assistant diagnostic system for breast cancer in which the landmarks are the mass density and the microcalcifications. It will be useful in analyzing mmmograms and screening for breast cancer. Less
|
