The claßification of blood cell via contrast-enhanced microholography and deep learning

Human blood analysis has provided rich information in rapid clinical diagnosis. Different from conventional blood cell counting method which is environment-dependent and costly, this study proposes an advanced blood cells imaging method at micron-scale to reduce the size of the equipment and decrease the total cost of testing. This approach applies the deep learning method and a convolutional neural network in reconstructing object images from the diffraction patterns. The holographic image is extracted by the convolution layer and the feature claßification of the hidden layer rapidly identifies each diffraction pattern of the holographic image. The mean IoU for masks generated from the hologram is 0.876. Consequently, this deep learning approach is significantly more preferable to conventional calculation. It, thus, provides a portable, compact and cost-effective contrast-enhanced microholography system for clinical diagnosis.