I. Introduction
Digital pathology and microscopy images play a significant role in decision making for disease diagnosis, since they can provide extensive information for CAD, which enables quantitative analysis of digital images with a high throughput processing rate. Nowadays, automatic digital pathology including image analysis, which can greatly benefit pathologists and patients, has attracted much attention in both research and clinical practice [1], [2]. In comparison with manual assessment that is labor intensive and time consuming, computerized approaches [3]–[6] provide faster and reproducible image analysis such that the basic science researchers and clinician scientists can be released from boring and repeated routine efforts. More importantly, the complex nature of pathology and microscopy images presents significant challenges for manual image analysis, which might lead to large interobserver variations [7]; on the other hand, the CAD can greatly reduce the bias and provide accurate characterization of diseases [8]. In addition, it allows personalized treatments that can significantly benefit the patients. In order to handle large-scale image datasets, grid computing [9]–[11] and computationally scalable algorithms [12]– [14] are reported for high-throughput pathology image analysis. Another advantage of automated methods is that they can easily provide reproducible and rigorous measurements of important image features, which will be used with clinical followup, and thus, allows comparative study and potential prognosis and personalized medicine.