Yoon W Jong, Brown Matthew A, Reinhall Per G, Park Sangtae, Seibel Eric J
Department of Mechanical and Industrial Engineering, Qatar University, Doha, Qatar.
Minim Invasive Ther Allied Technol. 2012 Sep;21(5):320-8. doi: 10.3109/13645706.2011.653374. Epub 2012 Feb 14.
The current gold standard of bladder cancer surveillance, endoscopic visualization, is manually manipulated and still has significant room for improvement in performance and controls.
This paper reports our developments toward automated bladder surveillance that employs a shape memory alloy-based machine-controlled scanning mechanism. In conjunction with the electro-mechanical advances, we use modified commercial post-processing computer vision software capable of converting cystoscopic video of the bladder into stitched panoramas.
Experimental results conducted on a synthetic bladder demonstrate that this computer-aided scanning tool can help 82% of the entire bladder surface being scanned. Although the panoramic stitching algorithm increases the field of view and generates reasonable results in many cases, some image matching failures result in incompleteness in its full panoramic reconstruction.
Our current study ensures that the automated steering mechanism can follow the desired trajectory to scan the surface of the bladder but must be improved. The current reconstruction algorithm needs further modification. Our methodology may constitute a first step in suggesting a new automated and computer-aided bladder surveillance system.
膀胱癌监测的当前金标准——内镜可视化，是人工操作的，在性能和控制方面仍有很大的改进空间。
本文报告了我们在基于形状记忆合金的机器控制扫描机制的自动膀胱监测方面的进展。结合机电方面的进展，我们使用经过修改的商业后处理计算机视觉软件，该软件能够将膀胱的膀胱镜视频转换为拼接全景图。
在合成膀胱上进行的实验结果表明，这种计算机辅助扫描工具可以帮助扫描整个膀胱表面的82%。尽管全景拼接算法增加了视野并在许多情况下产生了合理的结果，但一些图像匹配失败导致其完整全景重建不完整。
我们目前的研究确保了自动转向机制可以沿着所需轨迹扫描膀胱表面，但仍需改进。当前的重建算法需要进一步修改。我们的方法可能是提出一种新的自动和计算机辅助膀胱监测系统的第一步。
