Basic research on individual identification by dental information matrix and deep learning
| Project/Area Number |
17K12014
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Social dentistry
|
|---|
| Research Institution | Nagasaki University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2017-04-01 – 2021-03-31
|
| Project Status |
Completed (Fiscal Year 2020)
|
| Budget Amount *help |
¥4,680,000 (Direct Cost: ¥3,600,000、Indirect Cost: ¥1,080,000)
Fiscal Year 2019: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2018: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2017: ¥2,470,000 (Direct Cost: ¥1,900,000、Indirect Cost: ¥570,000)
|
|---|
| Keywords | 歯科医療管理 / 個人識別 / 深層学習 / deep learning / 口腔情報 / デンタルチャート / 大規模災害 / 身元確認 / 口腔内情報 / 歯式 / 歯科医療管理学 |
|---|
| Outline of Final Research Achievements |
In recent years, it has been recognized in forensic dentistry that dental information is effective for personal identification to unidentified person in the event of a large-scale disaster. In this study, basic research of individual identification by deep learning and evaluation of oral information matrix were investigated.
An attempt was made to automatically recognize the tooth classification by deep learning from the intraoral image, and the high recognition rate was obtained. Furthermore, it was revealed that the evidence for tooth classification was visualized by GradCAM and the anatomical features were captured by the neural network model. The oral information matrix definition adopted the oral examination information standard code system. However, since this code system produces variable-length data, it became clear that LSTM is suitable for neural networks.
|
| Academic Significance and Societal Importance of the Research Achievements |
本研究では口腔情報を表現する歯式マトリックスを定義し、深層学習で形成したニューラルネットワークに歯式マトリックスをパターン認識させて個人識別を行うものである。
現在、本院のデータウェアハウスを使用して、初診登録した患者に対して歯式情報で個人識別が可能となっているが、1 歯単位の検索のために個人特定に時間を要する。本研究により高速な個人識別が可能になると予想できる。大規模データベースで歯式マトリックスのニューラルネットワークによるパターン識別で個人識別が可能になれば、高速な検索が可能となることが予測され、東日本大震災のような大規模災害時の身元確認に貢献できる意義がある。
|
Report
(5 results)
Research Products
(15 results)
