MitoVis: A Unified Visual Analytics System for End-to-End Neuronal Mitochondria Analysis

Area of Science:

• Neuroscience
• Cell Biology
• Bioinformatics

Background:

• Mitochondrial morphology impacts neuronal function and neurodegenerative diseases.
• Current analysis methods are manual, time-consuming, and prone to errors.
• Deep learning tools exist but require fine-tuning and human proofreading for daily use.

Purpose of the Study:

• To introduce MitoVis, a visualization system for end-to-end analysis of neuronal mitochondria morphology.
• To develop an active learning framework for accurate deep learning model fine-tuning.
• To provide interactive visual guides for efficient error correction.

Main Methods:

• Developed MitoVis, an integrated visualization and analysis system.
• Implemented a novel active learning framework using contrastive learning.
• Incorporated interactive visual guides for proofreading deep learning model outputs.

Main Results:

• MitoVis enables accurate fine-tuning of neural network models for mitochondria analysis.
• Interactive proofreading significantly reduces the effort needed to correct errors.
• Achieved up to 13.3× faster total analysis time compared to manual workflows in case studies.

Conclusions:

• MitoVis offers an efficient and accurate solution for neuronal mitochondria morphology analysis.
• The system streamlines the workflow, making advanced analysis accessible for daily research.
• MitoVis aids neuroscientists in studying neuronal function and neurodegenerative diseases.