摘要
Segmentation of objects with known geometries in an image is a wide research area. In this paper we show an energy minimization model to detect the tip of glass pipettes in microscopy images. The described model fits two rectangles with a common reference point to dark image regions, which are the sides of a pipette. The model is minimized using gradient descent. The low number of parameters result in a fast evolution and noise insensitivity. The algorithm is tested on label-free and fluorescent microscopy images. The error of the tip detection is only a few micrometers. Automatic pipette tip detection is a step forward to automate the patch-clamping process. The described method can be extended to 3 dimensions or other applications.
摘要译文
在图像中分割具有已知几何形状的物体是一个广泛的研究领域。在本文中，我们展示了一种能量最小化模型来检测显微镜图像中玻璃移液管的尖端。所描述的模型将两个矩形配合到具有共同参考点的暗图像区域，它们是移液管的侧面。该模型使用梯度下降最小化。参数数量少导致快速演化和噪声不敏感性。该算法在无标签和荧光显微镜图像上进行测试。尖端检测的误差只有几微米。自动移液管尖端检测是自动化贴片夹持过程的一个步骤。所述方法可以扩展到3维或其他应用
Krisztian Koos1;József Molnár1;Peter Horvath12. Pipette Hunter: Patch-Clamp Pipette Detection. Image Analysis[M].DE: Springer;LNCS, 2017: 172-183