Towards Real-Time in Vitro cell Culture Monitoring: Detecting Changes in LN-18 Glial Cell Morphology Using a Supervised Machine Learning Approach

In cell-based research, the process of visually monitoring cell cultures generates large image datasets that need to be evaluated for quantifiable information in order to track effectiveness of treatments in vitro. However, the traditional approach to annotating image data is tedious and error prone. This project sought to create an image classification framework that employs machine learning to detect different LN-18 glial morphology. Cell images captured from the EVOL-L microscope system underwent preprocessing to enhance contrast and reduce noise. A training set was created using labeled images and extracted shape and texture features. A non-linear SVM with a Gaussian Radial Basis kernel was trained using the training set. The resulting classifier was tested using a test set containing new images. The testing period showed that the classifier had a classification accuracy of 76%. The trained classifier was then applied to a database containing 636 cell images.