@misc{oai:repo.qst.go.jp:00084680,
 author = {Ryusuke, Hirai and Yukinobu, Sakata and Shinichiro, Mori and Ryusuke, Hirai and Yukinobu, Sakata and Shinichiro, Mori},
 month = {Sep},
 note = {Purpose/Objective(s)

To achieve high treatment accuracy in respiratory-gated radiotherapy, it is necessary to obtain a real-time tumor position in 3D space with two fluoroscopic images from different directions. Unfortunately, the low contrast of fluoroscopy images makes tumor position tracking difficult. Several works approached the problem by tracking fiducial markers implanted near the tumor instead of tracking the tumor directly. Linear fiducial markers are easier to implant than spherical markers but much more difficult to track because marker rotation may produce different projection marker shapes in each image frame. Our objective is to develop real-time linear fiducial marker tracking with fluoroscopic images.

Materials/Methods

Before treatment, a medical physicist selected a suitable marker angle while checking the marker rotation in fluoroscopic images. Then masks specifying whether a given region is a marker region or not a marker region were made. Each mask included one rectangle that rotated near the selected angle. The inside of the rectangle was specified as the marker region on the mask, since the marker shape projected onto the image would be similar to the rectangle. The masks were set on a fluoroscopic image, and then each pixel under the mask was divided into a marker-class or not a marker-class and the separation metric, which is the ratio of within-class variance to between-class variance, was calculated. The maximum value of the metric for all masks was used as a marker likelihood at the position where the masks were set. For tracking the marker, a particle filter consisting of two steps was used. In the first step, the marker position was predicted from the past marker position. In the second step, the marker likelihood of each predicted position was calculated and the marker position was determined with the likelihood-weighted average of the positions. The method was evaluated using three liver fluoroscopic image datasets with an implanted linear fiducial marker 0.5 mm in diameter by 3.0 mm in length. Each dataset consisted of at least one respiratory phase. 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 on a workstation.

Results

The mean positional error for three patients was measured to be 0.41 mm with a standard deviation of 0.20 mm. The worst positional error was 1.35 mm. For the first, second and third patient, the mean errors were 0.31, 0.35 and 0.46 mm with a standard deviation of 0.16, 0.18 and 0.29 mm, respectively. The calculation time was within 24.0 milliseconds per frame.

Conclusion

We presented real-time linear fiducial marker tracking for respiratory-gated radiotherapy. The high accuracy and short computation time of the proposed method were shown in an experiment involving three liver patients., The Amercian Society for Radiation Oncology (ASTRO) Annual meeting 2019},
 title = {Real-time Linear Fiducial Marker Tracking in Respiratory-gated Radiotherapy for Hepatocellular Carcinoma},
 year = {2019}
}