1992 Fiscal Year Final Research Report Summary
Supporting system for the diagnosis of carcinoma based on ultrasonographs using a fuzzy inference
| Project/Area Number |
03808005
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Informatics
|
|---|
| Research Institution | Kawasaki Medical School |
|---|
Principal Investigator |
ARITA Seizaburo Kawasaki Medical School Dept.of Mathematics Associate Professor, 医学部, 助教授 (20098601)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
FUJIOKA Tomoaki Iwate Medical University Dept.of Urology Associate Professor, 医学部, 助教授 (80173409)
SAITO Taiichi Kawasaki Medical School Dept.of Pharmacology Professor, 医学部, 教授 (00048258)
|
|---|
| Project Period (FY) |
1991 – 1992
|
| Keywords | Fuzzy Inference / Diagnostic System / Ultrasonograph / Carcinoma / Computer System / Fuzzy Information / Medical Education / Medical Images |
|---|
| Research Abstract |
We have developed a computer diagnostic system of prostate cancer using fuzzy theory, which allows objective assessment and gives consistent diagnosis even if performed by a less-experienced urologist. The Japanese Urological Association and the Japanese Society of Ultrasonics in Medicine have published "General Rules for Ultrasonic Diagnosis of Prostatic Cancer". We used 5 of their 7 criteria : shape of the plane, architecture of the shape, capsule echogenicity, similarity of the planes, and internal echogenicity. To these, we added two items, identification of a hypoechoic lesion and shift of the postero-basal margin. A fuzzy scale was used to estimate the degree of presence of feature analogically, and fuzzy theory was adopted for the diagnostic logic. Two Membership functions (the input and output) were prepared for each item and the final assessment was made by the center of gravity of the max-sets of the output membership function. The data regarding each criterion was input onto a fuzzy scale on a computer display, using a mouse key. In our clinical trial with this computer system, there was only one false-negative case and two cases in the borderline zone among 34 untreated patients with prostate cancer. There were two false positive case and four cases in the borderline zone among the 43 patients with benign prostatic hyperplasia. Therefore, sensitivity was 96.2% and specificity was 88.5% in our series. Our diagnostic system appears to be beneficical for medical education.
|
