Neural Networks for Cell Identification, Classification and Tracking

The effective classification and tracking of cells obtained from modern staining techniques has significant limitations due to the necessity of training and utilizing an expert in the field who must manually identify each cell in each slide. Often times these slides are filled with “noise” cells that are not of particular interest to the researcher. The use of computation methods has the ability to effectively and efficiently enhance image quality, as well as identify and track target cell types over large data sets. Slides obtained from various staining techniques can be segmented and enhanced in order to counter the varying degrees of pixel intensity obtained throughout the staining process. The collection of characteristics obtained from each cell then allows for the accurate removal of cells that are not relevant to the current research. This process alone could greatly enhance the ability of researchers to obtain information from large batches of cell stain data, however the use of neural networks to identify and track target cells can alleviate the need of manual classification altogether. Utilizing slides containing only target cells neural networks can then be trained to successfully identify and track the movement and morphological changes of target cells. The use of computation techniques allows for the detection of large-scale trends and subtle changes to cell behavior that would not otherwise be possible. The information that is then gained can be used to more effectively understand and potentially target cells. The use of neural networks as well as computation image enhancement techniques has the potential to gather information and trends regarding target cells more efficiently and accurately than current manual techniques.