Visual Detection of Cells in Brain Tissue Slice for Patch Clamp System

Whole-cell patch clamping is the gold-standard technology for electrical activities recordings of living single cells. Since it is difficult to train patch clamp operators, automatic patch clamp technologies have gained much attention. The key step in fully automated patch clamping is visual detection of cells in brain tissue slice under view of differential interference contrast (DIC) microscope. This paper proposes a cell detection method based on DIC images, which includes image classification based on image clarity and visual detection of cells in brain tissue slice based on machine learning. Firstly, combining the gray features and texture features of DIC images, we define a clarity evaluation function based on principal component analysis (PCA). Afterwards, by using K-Means Algorithm, DIC images can be divided into clear class and blurred class. The precision of this method is 99.1% in test set, which indicates that it can select the images with clear cell contour from a series of DIC images for subsequent processing. Secondly, we design a neuron detection process based on sliding window and pre-trained classifier to take cell detection in the selected clear images. We finally choose GoogLeNet as classifier in the detection process after comparing the advantages and disadvantages of classification methods based on support vector machine (SVM) and deep learning. The results show that the detected cells are basically operable living cells. The proposed cell detection technology in this paper provides an effective visual tool for fully automated patch clamping.