Plantorgan hunter: a deep learning-based framework for quantitative profiling plant subcellular morphology

Abstract

Abstract Accurate delineation of plant cell organelles from electron microscope images is essential to understand subcellular behaviors and functions. Here, we develop a deep learning pipeline, organelle segmentation network (OrgSegNet) for pixel-wise segmentation to identify chloroplasts, mitochondria, nuclei, and vacuoles. OrgSegNet was evaluated on a large manually-annotated dataset of 6371 organelles collected from 13 plant species, and achieved a state-of-the-art segmentation performance of these organelles. To generalize the morphological characteristics of plant organelles, we defined three morphological metrics (shape-complexity, electron-density, and area), and released an open-source web tool “Plantorgan Hunter” allowing quantitative profiling of subcellular morphology. The functionalities of Plantorgan Hunter can be easily operated, and we believe that it will increase the efficiency and productivity of plant subcellular morphological characteristics for the plant science community.

Plant phenotyping relevance

植物細胞小器官の画像セグメンテーションと形態指標抽出を開発・評価し、データセットと公開ツールまで提供する、明確な植物フェノタイピング手法研究。

Code and data availability

The paper publicly releases its manually-annotated TEM organelle dataset (Science Data Bank), the OrgSegNet code and trained models (GitHub), and a web tool (cropopen.com) for quantitative subcellular morphology profiling.