1996 Fiscal Year Final Research Report Summary
Computer-assisted Diagnostic System for Ultrasonography Using Fuzzy Reasoning
| Project/Area Number |
07672050
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
病態科学系歯学(含放射線系歯学)
|
|---|
| Research Institution | Kyushu University |
|---|
Principal Investigator |
TANAKA Takemasa Fac.Dentistry, Dept.Oral and Maxillofacial Radiology Kyushu University Research Associate, 歯学部, 助手 (30163538)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
TOYOFUKU Fukai School of Health Sciences Kyushu University Associate Professor, 医療技術短期大学部, 助教授 (10117179)
TOKUMORI Kenji Fac.Dentistry, Dept.Oral and Maxillofacial Radiology Kyushu University Research, 歯学部, 助手 (40253463)
MIWA Kunihiro Fac.Dentistry, Dept.Oral and Maxillofacial Radiology Kyushu University Research, 歯学部, 助手 (10136509)
|
|---|
| Project Period (FY) |
1995 – 1996
|
| Keywords | fuzzy reasoning / computer-assisted diagnosis / expert system / ultrasonography |
|---|
| Research Abstract |
We have developed the computer-assisted diagnostic system for ultrasonography using fuzzy reasoning. The system was constructed to differentiate the metastatic lymph node from inflammatory ones and applied to the unskilled clinicians who were from one to three years after the graduation. To construct the system, we set up three fuzzy production rules according to the diagnostic criteria. We used max-min product composition method in fuzzy reasoning and centroid method for standard defuzzification. Furthermore, in order to simulate the experienced clinician's consideration, we referred the method of analytic hierarchy process (AHP). Consequently, the diagnostic accuracy, sensitivity, and specificity were improved.
We applied some kinds of defuzzification method to the computer-assisted diagnostic system and compared the diagnostic accuracy in each application. Centroid, bisector and middle of maximum method were thought to be excellent defuzzification method for our computer-assisted diagnostic system. On the other hand, largest of maximum method was superior to any other defuzzification for the diagnostic accuracy, and smallest of maximum method was superior to the diagnostic specificity. Therefore, it was suggested that the defuzzification method should be considered and selected for clinical requirement.
Moreover, we considered the relationship between the unaided score of the observer and the effect of diagnostic aid. It was indicated that these were in the intensive negative correlation and the relationship could be approximated to linear regression. It was suggested that our system was especially useful for the observers whose unaided scores were very low.
|
