Protein crystal state discrimination method by extracting line features on the image

In recent years, several automated crystallization systems are developing for realizing high-throughput X-ray structure analysis. However, evaluation process of growth state of protein crystallization samples is not automated yet. This paper proposed a method of state discrimination for crystallization samples experiments. Especially its target is to discriminate between diffractable extract (crystal) and the others. Human experts observe the shape and size of the extracts to evaluate the growth state of the sample. Therefore, we utilize to extract line feature from the image of the sample, and to discriminate the state by using discriminant analysis as a classifier based on line features. Due to verify the performance of the proposal method, we experimented to discriminate using the images which are taken by an automated crystallization system; TERA. As a result, correction ratio was over 80%.