Bone Marrow Cell Counting Method Based on Fourier Ptychographic Microscopy and Convolutional Neural Network

In bone marrow examination, the number of bone marrow cells is an essential parameter to judge the degree of myeloproliferative. In this paper, we propose a new bone marrow cell counting method based on Fourier ptychographic microscopy and convolutional neural network. We use Fourier ptychographic microscopy technology to obtain the intensity and phase images of bone marrow cells at first. Then, we combine the intensity and the phase image correspondingly to obtain a dual-channel image. We use the convolutional neural network to extract the characteristics of bone marrow cells in the dual-channel image, which can generate a density map. The number of bone marrow cells is realized by integrating the density map. The experimental results show that both the mean absolute error (MAE = 0.66) and mean square error (MSE = 0.67) of our method are lower than those existing methods.