Detection Of Avian Embryos Viability Based On Heartbeat Signal Analysis

Avian influenza vaccines are generally prepared using egg embryo culture methods.Dead egg embryos need to be removed in time,otherwise bacteria will be bred,causing pollution and affecting the quality of the vaccine.Therefore,during the production process of avian influenza vaccine,egg embryos need to be tested for viability multiple times to eliminate dead egg embryos.At present,the viability of egg embryos is mainly detected by manual biopsy method in China.This method requires a lot of labor,high detection cost,low efficiency,and it is difficult to guarantee the accuracy of detection.The modern vaccine industry urgently needs an accurate,efficient and low-cost automatic detection method for egg embryo viability.This thesis combines sequence classification and deep learning technology,and proposes a 9-day egg embryo viability detection method based on heartbeat signal analysis.Firstly,a fixed-length egg embryo heartbeat signal was collected with the Photo Plethysmo Graphy,and a digital filter was designed to filter out signal noise and make a data set.Then the time-series convolutional network(TCN)is used to classify the heartbeat signals.A TCN enhanced network with a two-branch structure is proposed.One of the branches is a basic TCN network,which is used to extract the contour shapes of the heartbeat signals of dead and live embryos.The other branch is an enhancement module composed of gated recurrent unit network,which is used to extract the timing characteristics of the heartbeat signal.The egg embryo heartbeat data was input to two branches at the same time,and the output features of the two branches were fused using vector stitching.Finally,the softmax classifier was used to output the network classification results.The experimental results show that the TCN enhanced network proposed in this thesis has excellent performance in the task of detecting egg embryo viability,and its detection accuracy rate is as high as 99.87%.At the same time,the three evaluation indicators of accuracy,recall and F1 score reached 0.9970,0.9987,and 0.9978,respectively,and can accurately judge the viability of egg embryos at the vaccine production site.