Pancreas++: Automated quantification of pancreatic islet cells in microscopy images

Abstract

The microscopic image analysis of pancreatic Islet of Langerhans morphology is crucial for the investigation of diabetes and metabolic diseases. Besides the general size of the islet, the percentage and relative position of glucagon-containing alpha-, and insulin-containing beta-cells is also important for pathophysiological analyses, especially in rodents. Hence, the ability to identify, quantify and spatially locate peripheral, and "involuted" alpha-cells in the islet core is an important analytical goal. There is a dearth of software available for the automated and sophisticated positional quantification of multiple cell types in the islet core. Manual analytical methods for these analyses, while relatively accurate, can suffer from a slow throughput rate as well as user-based biases. Here we describe a newly devel-oped pancreatic islet analytical software program, Pancreas++, which facilitates the fully automated, non-biased, and highly reproducible investigation of islet area and alpha-and beta-cell quantity as well as position within the islet for either single or large batches of fluorescent images. We demonstrate the utility and accuracy of Pancreas++ by compar-ing its performance to other pancreatic islet size and cell type (alpha, beta) quantification methods. Our Pancreas++ analysis was significantly faster than other methods, while still retaining low error rates and a high degree of result correlation with the manually generated reference standard. © 2013 Chen, Martin, Cai, Fiori, Egan, Siddiqui and Maudsley.