WEKO3
アイテム
Template-Less Real-Time Robust Marker Tracking for Respiratory-Gated Radiation Therapy
https://repo.qst.go.jp/records/84681
https://repo.qst.go.jp/records/84681fb67e723-9140-4e8a-a99d-704bb8563e80
| Item type | 会議発表用資料 / Presentation(1) | |||||
|---|---|---|---|---|---|---|
| 公開日 | 2021-11-12 | |||||
| タイトル | ||||||
| タイトル | Template-Less Real-Time Robust Marker Tracking for Respiratory-Gated Radiation Therapy | |||||
| 言語 | ||||||
| 言語 | eng | |||||
| 資源タイプ | ||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_c94f | |||||
| 資源タイプ | conference object | |||||
| アクセス権 | ||||||
| アクセス権 | metadata only access | |||||
| アクセス権URI | http://purl.org/coar/access_right/c_14cb | |||||
| 著者 |
Yasunori, Taguchi
× Yasunori, Taguchi× Ryusuke, Hirai× Yukinobu, Sakata× Shinichiro, Mori× Yasunori, Taguchi× Ryusuke, Hirai× Yukinobu, Sakata× Shinichiro, Mori |
|||||
| 抄録 | ||||||
| 内容記述タイプ | Abstract | |||||
| 内容記述 | Purpose/Objective(s) To achieve high treatment accuracy in respiratory-gated radiation therapy, it is necessary for fluoroscopic images from 2 different directions to obtain the tumor position in 3-dimensional (3D) space. Several publications approached the problem by tracking the fiducial markers, since tumors are not always clearly visible in fluoroscopic images. In these methods, since template matching based on zero-mean normalized cross-correlation (ZNCC) is used, registration of template image is required. ZNCC is robust for linear brightness variations, however not for texture variations due to a patient’s motion and/or image noise. Our purpose is to develop a real-time robust fiducial marker tracking without template matching. Materials/Methods For tracking the marker, a particle filter consisting of 2 steps was used. In the first step, the marker position was predicted from the past marker trajectory. In the second step, the marker likelihood of each position near the predicted position was calculated and the marker position was detected as the one with largest likelihood. The likelihood was defined as the correlation ratio between 2 brightness distributions inside and outside a region that is determined according to the shape of the marker. In the case that the shape of the marker is spherical, the region is defined to be circular. The correlation ratio can be considered as the degree of separation between the 2 distributions. For evaluation, 6 lung cancer patients’ images including 18 spherical markers were used. Two flat-panel detectors were used, each with a resolution of 768 by 768 pixels. Pixel spacing was 0.388 mm/pixel. On the image plane including the isocenter, pixel spacing was 0.273 mm/pixel. The positional error of each marker was calculated between the tracked position and ground truth position in 3D space. In addition, the calculation time was measured. For comparison, a different method was also used for the same task. In the second method, the proposed correlation ratio was replaced by standard ZNCC. Results All markers were tracked successfully by the proposed method, whereas the ZNCC-based method lost the trajectory when texture varied. The mean positional error of the proposed method was measured to be 0.280 mm with a standard deviation of 0.109 mm. The calculation time was 3 milliseconds per frame. |
|||||
| 会議概要（会議名, 開催地, 会期, 主催者等） | ||||||
| 内容記述タイプ | Other | |||||
| 内容記述 | The American Society for Radiation Oncology (ASTRO) Annual meeting 2016 | |||||
| 発表年月日 | ||||||
| 日付 | 2016-09-26 | |||||
| 日付タイプ | Issued | |||||
