Mitochondria segmentation in electron microscopy volumes using deep convolutional neural network

Abstract

Mitochondria are organelles that play an important role in the cell's life cycle as the energy generating units. State-of-the-art imaging modalities, such as electron microscopy, allow researchers to study tissues, cells and sub-cellular organelles at high resolution. Recently, various works address the problem of segmenting mitochondria in electron microscopy images. Manual segmentation of mitochondria is difficult and may not have high accuracy as automatic segmentation can yield. In this paper, we present a deep convolutional neural network approach for automatic segmentation of mitochondria in brain tissue, specifically the CA1 hippocampus region imaged by an focusedion beam scanning electron microscope. The performance of the proposed method has been quantitatively evaluated. According to our experiments, deep convolutional neural network is a suitable solution for mitochondria segmentation. Results have been compared with previous studies that segment the mitochondria on CA1 Hippocampus Dataset. The proposed deep learning system produces promising results for segmentation of electron microscopy images.

Ähnliche Arbeiten