Neural network scoring of spots in X-Gal and -leu plates

We have developed an image analysis system for scoring yeast growth and color development in images of 96-well plates, a common format for high throughput assays. We use a segmentation method to locate the plates and spots. Color histogram and wavelet features are extracted respectively from spots of X-Gal and -leu plates. Two artificial neural networks are separately employed to score spots on each plate. The performance of the system is evaluated using a data set of 50 images. The data set was divided into 25 training and 25 testing images. Accuracies of 99.7% and 95.2% have been achieved for scoring the X-Gal and -leu plates respectively.