2019 Fiscal Year Final Research Report
Constructing a system for evaluating swallowing function by combination of swallowing sounds with neck depth images
| Project/Area Number |
17K01571
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Rehabilitation science/Welfare engineering
|
|---|
| Research Institution | The University of Shiga Prefecture |
|---|
Principal Investigator |
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
小澤 惠子 滋賀県立大学, 人間文化学部, 准教授 (90747429)
|
|---|
| Project Period (FY) |
2017-04-01 – 2020-03-31
|
| Keywords | 嚥下評価 / 嚥下音 / 深度画像 / 機械学習 / SVM / 頸部 |
|---|
| Outline of Final Research Achievements |
A SVM based classifier for evaluating the level of swallowing disorder could be constructed by using the swallowing sounds annotated of the VE score by clinicians as training data that was collected from healthy subjects and patients. The performance achieved to 78% accuracy in the two class problems for screening, and 46% accuracy in the four class problems in which swallowing status was categorized into four levels.
A noble algorithm could be successfully constructed as a non-contact technique for estimating swallowing time periods by using depth images of neck part. Applying the developed algorithm to the neck part depth image from both male and female healthy subjects yielded the 90% accuracy of the estimated swallowing periods.
|
| Free Research Field |
信号処理
|
| Academic Significance and Societal Importance of the Research Achievements |
言語聴覚士による嚥下評価では触診や聴診を行うが、今回開発した深度画像による嚥下時間推定手法を用いると、嚥下の行われたタイミングを触診無しで行い得ることができる。嚥下状態の遠隔診断を確立するために必要な一つの要素技術として、嚥下のタイミングを検知することは重要であるが、そのための一つの手法となり得る。一方、嚥下音の解析は精度の点ではまだ十分ではないが、スクリーニングだけでなく、嚥下の状態に応じたクラス識別についての可能性は少なくとも確認できた。このことから、遠隔診断だけでなく、在宅療養やケアハウスといった看護や介護の現場における簡易型嚥下評価システム構築の可能性を明らかにすることができた。
|
