An Accurate And Efficient Software For Detection Of Types Of Muscle Fiber Section

Pigs and foods are living and working in peace and contentment,pork safety is not only a millennium heritage of Chinese farming civilization,but also an indispensable "fixing man" on our journey towards a comprehensive well-off.With the continuous improvement of people’s living standards,people not only have to eat,but also eat meat and egg milk,not only meat,but also healthy meat and high-quality meat.The phenotypic standards for measuring meat quality mainly include PH value,dripping,meat color,and marbling.With the deepening of research,the research on meat quality has also progressed from macro to micro.The type of muscle fiber is a table that is highly related to meat quality at the cell level.Type,accurate measurement,easy to understand and analyze physiological and biochemical functions,has become one of the important means of pork quality improvement.However,the statistics of muscle fiber types need to be observed and counted after staining the muscle cell slices.In order to count accurately,a sample is generally counted 3 times,and the microscopic field of view of each slice under the microscope is repeated 5 times.Manual statistics are extremely time-consuming It is labor-intensive and error-prone,which greatly limits the development of this field.The rapid development of imaging technology has brought opportunities for the calculation of muscle fiber types.The image classification system built based on machine learning method can quickly segment the image area and count the area of the area,which is extremely suitable for the recognition and counting of different muscle cells after muscle fiber sectioning.This study intends to use machine learning and traditional image recognition algorithms to identify and count different ATPase-stained pig muscle fiber slices to provide highly automated and highly accurate software and web pages for the automatic identification,typing,and statistics of muscle fiber slices.Among them,the identification of muscle fiber pictures is completed in two steps.The first is to use machine learning to identify type I and type II muscle fiber cells.We randomly took 40 from 209 muscle fiber slices.For type I and type II Cells are identified(foreground color and background color)and trained using the pixel segmentation machine learning algorithm of ilastik software to construct a feature database of different types of cells.Then the remaining 169 pictures are predicted through this kind of database to obtain the number and area of type I cells and type II cells.Comparing the calculation results with the manual calculation data,it was found that the correlation of type I cells is 0.907 and the correlation of type II cell recognition is 0.9028,indicating that the software has extremely high classification accuracy for the two types of cells.The second step is to further classify type II muscle fiber cells into type IIA cells and type IIB cells.Since the type IIA and type IIB cells are not significantly different in the RGB color gamut,this study converted the RGB color gamut to the HSV color gamut After the space,the clustering algorithm was used to type IIA cells and type IIB cells.The recognition correlation between type IIA and type IIB cells was 0.800 and 0.861,which also proved the reliability of type II cell classification.In this study,the characteristic database of type I and type II cells of muscle fiber sections stained with pig ATPase was constructed.Using this database,the cell types and numbers of pig muscle fiber sections can be automatically identified,counted,and identified with an accuracy of up to 90%.This study also realized the automatic identification and identification of type IIA and type IIB cells,and the identification correlation was more than80%;after automating the two-step identification and calculation process,we wrote a software for automatic classification and counting of muscle fibers;At the same time,in order to facilitate the use of scientific research workers without programming background,we have also built an automated analysis website:http://www.jxau.edu.cn/210.35.128.207/home/zhiyan/Cellidentification/web/index.htm,users only need to upload ATPase stained muscle fiber slices,and the server will automatically return the number of different cells.Area and other result parameters.