Original images of colloidal gold immunochromatographic strip
Accurate segmentation of test line and control line for colloidal gold immunochromatographic strip (GICS) images with image processing algorithms is essential to quantitative analysis of GICS. As common methods for GICS image segmentation, fuzzy c-means (FCM) algorithm and cellular neural network (CNN) algorithm both require presetting initial conditions (specifying initial parameters or training models) and take long running time, due to high calculation cost. Therefore neither is ideal for a point-of-care testing (POCT) device, which has low hardware cost and limited computing power. This paper designs a region growing algorithm combined with fast peak detection (RGFPD) to quickly and self-adaptively segment GICS images. Compared with FCM algorithm and CNN algorithm, the RGFPD algorithm has two obvious advantages. First, the region growing algorithm requires low calculation cost, which better suits the POCT device, because it is a local algorithm, rather than a global one. Second, the fast peak detection algorithm calculates the seed points and growing criterions as initial conditions, realizing self-adaptive segmentation of images. In this paper, RGFPD algorithm is applied to segment GICS images of actual samples, taking FCM algorithm and CNN algorithm as contrast. The results show that RGFPD accurately segment images without presetting initial conditions, with shorter algorithm running time, and performs better in anti-interference.
The dataset contain some GICS images.