2014 Volume 22 Pages 1-7
Recent advances in bioimaging equipment have allowed for the significant and rapid acquisition of biological images. Consequently, new methods are energetically being developed to quantitatively and objectively evaluate the biological features of many bioimages. Recently, we described the application of the CARTA (Clustering-Aided Rapid Training Agent) image classification framework, which can be used to obtain highly accurate classifications of a wide variety of biomedical images. However, the efficacy of this technology in association with transmission electron microscopy (TEM) images has not been evaluated. We herein report the automatic classification of chloroplast TEM images of chloroplasts of wild-type Arabidopsis thaliana and four kinds of mutants impaired in chloroplast-targeted proteins. CARTA successfully classified the TEM images based on their genetic backgrounds. As previously reported with fluorescent microscopic and magnetic resonance images, CARTA reduced the annotation cost while maintaining accuracy. Furthermore, using images obtained with different accelerating voltages resulted in biologically meaningless classifications, suggesting that TEM settings are critical for chloroplast morphological classification. CARTA is expected to be useful for future computational inspections of the ultrastructure visualized with TEM.