| Stem cells are the original cells of living organisms,which have the ability of self-renewal and multi-directional differentiation potential.They can differentiate into many kinds of tissue cells of living organisms,and have important research value.At present,scholars at home and abroad have done a lot of research work on stem cells and achieved fruitful results.Some research results show that the functional characteristics of stem cells not only depend on their own biological structures,but also highly depend on the current physiological state of the cells such as cell viability.Therefore,it is of great significance to obtain real-time and accurate physiological state information of stem cells for their culture in vitro,drug development and clinical application.At present,the traditional methods,such as biochemistry,morphology and immunology,are commonly used to detect cell activity.However,these methods are often invasive.They will change the normal growth environment and physiological function of cells,and thus cause irreversible damage or death to the cells.In order to solve the shortcomings,we proposed a new way of rapid and non-destructive detection of human Umbilical Cord Mesenchymal Stem Cells(h UC-MSCs)viability by combining three-dimensional(3D)fluorescence technology with deep learning.The main research work of this paper is as follows:(1)The fluorescence self-quenching effect of mesenchymal stem cells was studied,and the optimal excitation wavelength and the fluorescence emission wavelength range were determined.The results showed that the concentration of h UC-MSCs in the range of 2x10~4?2x10~5/ml and the excitation wavelength range of240-300nm showed a good linear relationship between the concentration of h UC-MSCs and the fluorescence intensity,which satisfied the principle of fluorescence additivity.At a cell concentration of 1x10~5/ml,the 3D fluorescence spectra of h UC-MSCs with different viability were collected and analyzed.Based on the principle of fluorescence additivity,the 3D fluorescence spectra of h UC-MSCs with the same viability state are randomly weighted and superimposed to generate a new 3D fluorescence spectrum,which is used to establish a deep learning training data set.(2)According to the traditional VGGNet neural network,a network model VGGNet-10 suitable for 3D fluorescence spectral analysis of h UC-MSCs was proposed by optimizing the gradient descent algorithm and adjusting a series of hyperparameters used in the network.As a result,the network has a simple structure,a strong network generalization ability,and the low requirements of computer performance.In addition,The network can obtain the test results of h UC-MSCs viability in only 10 seconds,with an accuracy rate as high as 90.7%.(3)The algorithm of h UC-MSCs viability discrimination based on 3D fluorescence spectroscopy was studied,and a h UC-MSCs viability detection software platform was developed.Based on the Python language combined with the Tkinter visualization library,a visualization interface that can interact with the user was drawn.An analysis software based on 3D fluorescence spectroscopy to automatically detect the viability of h UC-MSCs was designed.The software platform integrates functions such as data conversion,network model training,and unknown cell viability detection,providing a convenient,fast,and accurate analysis tool for h UC-MSCs viability detection.To sum up,compared with the traditional methods of cell viability detection,the method proposed in this paper combines 3D fluorescence with deep learning has the characteristics of fast,unmarked and so on,which provides a new way to solve the problems faced by the traditional detection methods,such as invasion,damage to cell growth environment and so on.In summary,compared with the traditional cell viability detection method,the method proposed in this paper combines 3D fluorescence spectroscopy and deep learning,and thus provides a label-free,non-invasive and rapid method for h UC-MSCs viability assessment. |
PDF Full Text Request