| 研究課題/領域番号 |
16F16108
|
|---|
| 研究機関 | 九州大学 |
|---|
研究代表者 |
橋爪 誠 九州大学, 医学研究院, 教授 (90198664)
|
|---|
| 研究分担者 |
JANG JONGSEONG 九州大学, 医学(系)研究科(研究院), 外国人特別研究員
|
|---|
| 研究期間 (年度) |
2016-07-27 – 2019-03-31
|
| キーワード | Surgical navigation / Medical image analysis / Object tracking / Artificial Intelligence |
|---|
| 研究実績の概要 |
In this year, I archived in implementation of the components of the navigation for robot assisted radical prostectomy. I implemented an artificial intelligence based forceps tracking software. This software can track and show the forceps’ end points in the surgical movie by using the region proposal network (RPN). As a result, RPN showed its possibility to detect the forceps in the very complex laparoscopic surgical scenes. It was not yet enough to apply to the navigation system due to frame rate. Resizing the input image smaller can be a solution, because many algorithms for object detection have used input image smaller than 600 × 600 at least. However, there is trade-off between accuracy and frame rate when using smaller image, it has to be investigated through experiments.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
3: やや遅れている
理由
There was no big problem in implementation of AI software. But, there were some difficulties in getting image data to train and test. Also, it took much time to train itself and check the its evaluation. The result was very promise, but it needs more data and time to get robust performance.
|
| 今後の研究の推進方策 |
I’ll continue to study in tracking forceps using AI. So far, obtained forceps’ position was the result of the image from the single camera, so that it is 2 D. It should be transformed to 3 D through the stereo vision technology using multi-cameras. Afterwards, it can be applied to the navigation system of robot surgical system as a principal component.
|
