Nellie: Automated organelle segmentation, tracking, and hierarchical feature extraction in 2D/3D live-cell microscopy

Abstract: The analysis of dynamic organelles remains a formidable challenge, though key to understanding biological processes. We introduce Nellie, an automated and unbiased user-friendly pipeline for segmentation, tracking, and feature extraction of diverse intracellular structures. Nellie adapts to image metadata, eliminating user input. Nellie's preprocessing pipeline enhances structural contrast on multiple intracellular scales allowing for robust hierarchical segmentation of sub-organellar regions. Internal motion capture markers are generated and tracked via a radius-adaptive pattern matching scheme, and used as guides for sub-voxel flow interpolation. Nellie extracts a plethora of features at multiple hierarchical levels for deep and customizable analysis. Nellie features a point-and-click Napari-based GUI that allows for code-free operation and visualization, while its modular open-source codebase invites extension by experienced users. We demonstrate Nellie's wide variety of use cases with three examples: unmixing multiple organelles from a single channel using feature-based classification, training an unsupervised graph autoencoder on mitochondrial multi-mesh graphs to quantify latent space embedding changes following ionomycin treatment, and performing in-depth characterization and comparison of endoplasmic reticulum networks across different cell types and temporal frames.

• The paper introduces a novel multi-scale preprocessing method using an adaptive modified Frangi filter to enhance organelle segmentation without manual input.
• It presents a robust tracking algorithm with motion capture markers and sub-voxel interpolation to accurately follow organelle dynamics in live cells.
• The study pioneers hierarchical feature extraction by analyzing structures from individual voxels to complex organelle networks, enabling deep intracellular insights.

Comprehensive Overview of Nellie: Automating Organelle Analysis in Live-Cell Microscopy

Introduction to Nellie

Nellie introduces a novel image analysis pipeline designed to automate the segmentation, tracking, and feature extraction of intracellular organelles across multiple scales in live-cell microscopy data. It distinguishes itself by adaptable preprocessing methods, enabling robust segmentation of organelles and their substructures without requiring manual input. Further innovation is seen in its tracking algorithm, which uses motion capture markers and sub-voxel flow interpolation, enhancing accuracy in tracking organelle dynamics. Nellie's feature extraction capabilities are extensive, gathering a comprehensive array of both standard and advanced metrics across multiple hierarchical levels of organellar organization.

Technological Advances and Methodological Strategies

Nellie employs several key advancements to address existing challenges in organelle analysis:

• Adaptive Multi-Scale Preprocessing: Leveraging a multi-scale modified Frangi filter, Nellie enhances structural contrast at varying intracellular scales. Its preprocessing is uniquely adaptive, adjusting filter parameters based on image metadata to maximize segmentation efficacy.
• Motion Capture and Tracking: Unlike traditional methods, Nellie generates internal motion capture markers through radius-adaptive pattern matching. This method greatly improves tracking accuracy in densely populated and dynamic cellular environments.
• Hierarchical Feature Extraction: With a new approach to dissect organelles into hierarchical levels, Nellie enables deep analysis by extracting features at every level – from single voxels to the entire cellular landscape. This allows for nuanced understanding of organelle behavior and organization.

Practical Implications and Future Prospects

Nellie was evaluated on its ability to demix fluorescent signals of different organelles in single-channel images and to construct multi-mesh organelle graphs for advanced analyses. These capabilities pave the way for unprecedented applications in cellular biology, including but not limited to, complex organelle network analysis and the study of intracellular dynamics under various conditions.

1. Unmixing Organelles in Single-Channel Images: Demonstrating its powerful feature extraction, Nellie successfully used morphological and motility features to distinguish between Golgi apparatus and mitochondria in mixed fluorescent images. This capability has significant implications for maximizing information retrieval from limited imaging channels.
2. Multi-Mesh Organelle Graphs: By adopting a novel multi-mesh approach, Nellie facilitated the construction of graph representations of organelle networks. This was exemplarily used to analyze mitochondrial network alterations, showcasing the potential for deep structural and functional insights into intracellular organization.

Conclusion

Nellie sets a new standard for automation and comprehensiveness in organelle analysis within live-cell microscopy. Its modular and user-friendly design, combined with the depth of analysis it offers, both empowers and inspires further exploration in the dynamic field of cellular biology. Future developments could see its application in broader contexts, bridging between cellular biochemistry, pathology, and beyond, contributing significantly to our understanding of cellular functions and diseases.

Ethical and Accessibility Considerations

Nellie has been made openly accessible, promoting transparency and fostering collaboration within the scientific community. The emphasis on ethical development and data handling underlines the commitment to responsible research and innovation.

Future Directions

Expanding Nellie's capabilities to include more organelle types and incorporating machine learning models for predictive analysis of organelle behavior are promising directions. Furthermore, integration with other cellular imaging techniques could broaden its application scope, making it an indispensable tool in cellular and molecular biology research.