An Online Learning Approach for Robust Motion Tracking in Liver Ultrasound Sequence
| Autor: | Huabei Shi, Chunxu Shen, Jian Wu, Yibin Huang, Tao Sun |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Computer science
business.industry Online learning Adaptive kernel Correlation filter Liver ultrasound 030218 nuclear medicine & medical imaging Tracking error 03 medical and health sciences 0302 clinical medicine Match moving Robustness (computer science) Computer vision Artificial intelligence Abdominal radiotherapy business 030217 neurology & neurosurgery |
| Zdroj: | Pattern Recognition and Computer Vision ISBN: 9783030033378 PRCV (3) |
| DOI: | 10.1007/978-3-030-03338-5_37 |
| Popis: | Suffering from respiratory motion and drift, radiotherapy requires real-time and accuracy motion tracking to minimize damage to critical structures and optimize dosage delivery to target. In this paper, we propose a robust tracker to minimize tracking error and enhance the quality of radiotherapy based on two-dimensional ultrasound sequences. We firstly develop a scale adaptive kernel correlation filter to compensate deformation. Then the filter with an improved update rule is utilized to predict target position. Moreover, displacement and appearance constrains are elaborately devised to restrict unreasonable positions. Finally, a weighted displacement is calculated to further improve the robustness. Proposed method has been evaluated on 53 targets, yielding 1.13 ± 1.07 mm mean and 2.31 mm 95%ile tracking error. Extensive experiments are performed between proposed and state-of-the-art algorithms, and results show our algorithm is more competitive. Favorable agreement between automatically and manually tracked displacements proves proposed algorithm has potential for target motion tracking in abdominal radiotherapy. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|