Leveraging genetic algorithm and neural network in automated protein crystal recognition.

We propose a classification framework combined with a multi-scale image processing method for recognizing protein crystals in high-throughput images. The main three points of the processing method are the multiple population genetic algorithm for region of interest detection, multi-scale Laplacian pyramid filters and histogram analysis techniques to find an effective feature vector. Using human (expert crystallographers) classified images as ground truth, the current experimental results gave 88% true positive and 99% true negative rates, resulting in an average true performance of approximately 93.5% validated on an image database which contained over 79,000 images.