Towards unsupervised segmentation in high-resolution medical nano-imaging

Abstract

Recent advances in cellular and subcellular microscopy demonstrated its potential towards unraveling the mechanisms of various diseases at the molecular level. From a computer vision perspective nano-imaging is an inherently complex environment as can for example be seen from Fig.1(a,c). For the image analysis of intracellular organisms in high-resolution microscopy, new techniques which are capable of handling high-throughput data in a single pass and real time are of special interest. The additional emphasis is put therein on automated solutions which can provide the objective quantitative information in a reasonable time frame. The state-of-the-art is dominated by manual data annotation[1]and the early attempts to automate the segmentation are based on statistical machine-learning techniques[4].