Efficient cell-wide mapping of mitochondria in electron microscopic volumes using webKnossos

Area of Science:

• Cell Biology
• Neuroscience
• Computational Biology

Background:

• Volume electron microscopy (vEM) and AI enable high-throughput cellular structure quantification.
• Assigning cell identity to segmented mitochondria is a computational challenge, often requiring cell membrane contouring.

Purpose of the Study:

• Introduce mito-SegEM, a vEM reconstruction procedure for assigning cell identity to segmented mitochondria.
• Bypass the need for cell membrane contouring in mitochondrial instance segmentation.

Main Methods:

• Utilize virtual-path-based annotation for mitochondrial instance assignment.
• Employ webKnossos, an open-source platform, for annotation, visualization, and proofreading.
• Apply the procedure to volumetric datasets from brain, intestine, and testis tissues.

Main Results:

• Successfully assigned cell identities to numerous mitochondrial instances across diverse tissues.
• Demonstrated accurate and efficient reconstruction of mitochondria.
• Validated the broad applicability of the mito-SegEM procedure.

Conclusions:

• Mito-SegEM offers an efficient solution for cellular-scale mitochondrial reconstruction.
• The method overcomes computational challenges in organelle instance segmentation.
• Facilitates advanced quantitative analysis of mitochondrial networks in various biological contexts.