<p>Needle-based procedures using robotic systems are widely used in minimally invasive surgeries for diagnostic and therapeutic purposes. To insert the needle accurately, precise measurement of the needle tip’s position and overall needle shape is essential. This paper introduces a robotic ultrasound tracking system that integrates an ultrasound probe with a robotic system and uses a U-Net-based needle detection algorithm to track and reconstruct needle shape accurately. Our system addresses the issue of comet-tail artifacts around metal needles in ultrasound images, by restoring the needle cross-section to a circular shape. In the experiments, the proposed system tracked the pre-inserted needle and recorded its position for the reconstruction of 3-dimensional needle shape. Our results show that needle shapes in transparent gelatin phantoms and in the transparent parts of phantoms containing opaque biological tissues can be tracked with root-mean-square errors of 0.279±0.115 mm and 0.430±0.193 mm, respectively. The proposed method is effective for evaluating the performance of needle insertion robots and its simplicity and improved performance promise highly accurate needle positioning and enhanced therapeutic outcomes.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Three-dimensional Needle Shape Reconstruction Using Robotic Ultrasound Tracking System With U-Net-based Segmentation Model

  • Jiye Lee,
  • Bijoya Lala,
  • Seong Young Ko

摘要

Needle-based procedures using robotic systems are widely used in minimally invasive surgeries for diagnostic and therapeutic purposes. To insert the needle accurately, precise measurement of the needle tip’s position and overall needle shape is essential. This paper introduces a robotic ultrasound tracking system that integrates an ultrasound probe with a robotic system and uses a U-Net-based needle detection algorithm to track and reconstruct needle shape accurately. Our system addresses the issue of comet-tail artifacts around metal needles in ultrasound images, by restoring the needle cross-section to a circular shape. In the experiments, the proposed system tracked the pre-inserted needle and recorded its position for the reconstruction of 3-dimensional needle shape. Our results show that needle shapes in transparent gelatin phantoms and in the transparent parts of phantoms containing opaque biological tissues can be tracked with root-mean-square errors of 0.279±0.115 mm and 0.430±0.193 mm, respectively. The proposed method is effective for evaluating the performance of needle insertion robots and its simplicity and improved performance promise highly accurate needle positioning and enhanced therapeutic outcomes.