Online Demo


The SHIRAZ Project

SHIRAZ is a collaborative project at Penn State University between the laboratories of James Z. Wang (College of Information Sciences and Technology) and Keith Cheng (College of Medicine). The project integrates knowledge of computer vision, machine learning, pathology, and genomics to produce a content-based image retrieval (CBIR) system for the automated characterization and high-throughput phenotypic annotation of animal histology images.

Project Summary

In order to elucidate the function of uncharacterized human genes, scientists analyze the phenotypes resulting from mutations or knockdowns of homologous genes in model organisms such as the mouse, fruit fly, and zebrafish. A thorough, systematic interpretation of the relationship of phenotypes with their molecular-level genotypes can strongly benefit from large-scale, high-throughput phenotyping studies. Histology, the microscopic study of biological tissues, is a highly sensitive method of characterizing subtle phenotypes. Recent technological advancements have enabled the largely automatic collection of high-throughput histological data in the form of digital image libraries. The task of scoring and annotating the collected images, however, remains a tedious and subjective process. In the proposed research, the investigators will apply novel methods in image processing and statistical learning to facilitate the rapid, quantitative characterization and automatic annotation of histological phenotypes, with a pilot application designed for zebrafish histology. It is anticipated that the proposed system will simultaneously advance the state of the art in biological image retrieval and facilitate the scientific community's use of histology as an effective high-throughput phenotyping method for functional genomics and systems biology.

Figure 1. An example of "frieze-like expansion" technology, a novel approach to feature
extraction and classification of histology images exhibiting implicit rotational symmetry.
(For more information, please see our CIVR 2008 paper and others in the Publications section.)