CardioVinci: building blocks for virtual cardiac cells using deep learning.

Advances in electron microscopy (EM) such as electron tomography and focused ion-beam scanning electron microscopy provide unprecedented, three-dimensional views of cardiac ultrastructures within sample volumes ranging from hundreds of nanometres to hundreds of micrometres. The datasets from these samples are typically large, with file sizes ranging from gigabytes to terabytes and the number of image slices within the three-dimensional stack in the hundreds. A significant bottleneck with these large datasets is the time taken to extract and statistically analyse three-dimensional changes in cardiac ultrastructures. This is because of the inherently low contrast and the significant amount of structural detail that is present in EM images. These datasets often require manual annotation, which needs substantial person-hours and may result in only partial segmentation that makes quantitative analysis of the three-dimensional volumes infeasible. We present CardioVinci, a deep learning workflow to automatically segment and statistically quantify the morphologies and spatial assembly of mitochondria, myofibrils and Z-discs with minimal manual annotation. The workflow encodes a probabilistic model of the three-dimensional cardiomyocyte using a generative adversarial network. This generative model can be used to create new models of cardiomyocyte architecture that reflect variations in morphologies and cell architecture found in EM datasets. This article is part of the theme issue 'The cardiomyocyte: new revelations on the interplay between architecture and function in growth, health, and disease'.

The Cell Physiome: What Do We Need in a Computational Physiology Framework for Predicting Single-Cell Biology?

Modern biology and biomedicine are undergoing a big data explosion, needing advanced computational algorithms to extract mechanistic insights on the physiological state of living cells. We present the motivation for the Cell Physiome Project: a framework and approach for creating, sharing, and using biophysics-based computational models of single-cell physiology. Using examples in calcium signaling, bioenergetics, and endosomal trafficking, we highlight the need for spatially detailed, biophysics-based computational models to uncover new mechanisms underlying cell biology. We review progress and challenges to date toward creating cell physiome models. We then introduce bond graphs as an efficient way to create cell physiome models that integrate chemical, mechanical, electromagnetic, and thermal processes while maintaining mass and energy balance. Bond graphs enhance modularization and reusability of computational models of cells at scale. We conclude with a look forward at steps that will help fully realize this exciting new field of mechanistic biomedical data science.

EM-stellar: benchmarking deep learning for electron microscopy image segmentation.

MOTIVATION: The inherent low contrast of electron microscopy (EM) datasets presents a significant challenge for rapid segmentation of cellular ultrastructures from EM data. This challenge is particularly prominent when working with high-resolution big-datasets that are now acquired using electron tomography and serial block-face imaging techniques. Deep learning (DL) methods offer an exciting opportunity to automate the segmentation process by learning from manual annotations of a small sample of EM data. While many DL methods are being rapidly adopted to segment EM data no benchmark analysis has been conducted on these methods to date. RESULTS: We present EM-stellar, a platform that is hosted on Google Colab that can be used to benchmark the performance of a range of state-of-the-art DL methods on user-provided datasets. Using EM-stellar we show that the performance of any DL method is dependent on the properties of the images being segmented. It also follows that no single DL method performs consistently across all performance evaluation metrics. AVAILABILITY AND IMPLEMENTATION: EM-stellar (code and data) is written in Python and is freely available under MIT license on GitHub (https://github.com/cellsmb/em-stellar). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Automated segmentation of cardiomyocyte Z-disks from high-throughput scanning electron microscopy data.

BACKGROUND: With the advent of new high-throughput electron microscopy techniques such as serial block-face scanning electron microscopy (SBF-SEM) and focused ion-beam scanning electron microscopy (FIB-SEM) biomedical scientists can study sub-cellular structural mechanisms of heart disease at high resolution and high volume. Among several key components that determine healthy contractile function in cardiomyocytes are Z-disks or Z-lines, which are located at the lateral borders of the sarcomere, the fundamental unit of striated muscle. Z-disks play the important role of anchoring contractile proteins within the cell that make the heartbeat. Changes to their organization can affect the force with which the cardiomyocyte contracts and may also affect signaling pathways that regulate cardiomyocyte health and function. Compared to other components in the cell, such as mitochondria, Z-disks appear as very thin linear structures in microscopy data with limited difference in contrast to the remaining components of the cell. METHODS: In this paper, we propose to generate a 3D model of Z-disks within single adult cardiac cells from an automated segmentation of a large serial-block-face scanning electron microscopy (SBF-SEM) dataset. The proposed fully automated segmentation scheme is comprised of three main modules including "pre-processing", "segmentation" and "refinement". We represent a simple, yet effective model to perform segmentation and refinement steps. Contrast stretching, and Gaussian kernels are used to pre-process the dataset, and well-known "Sobel operators" are used in the segmentation module. RESULTS: We have validated our model by comparing segmentation results with ground-truth annotated Z-disks in terms of pixel-wise accuracy. The results show that our model correctly detects Z-disks with 90.56% accuracy. We also compare and contrast the accuracy of the proposed algorithm in segmenting a FIB-SEM dataset against the accuracy of segmentations from a machine learning program called Ilastik and discuss the advantages and disadvantages that these two approaches have. CONCLUSIONS: Our validation results demonstrate the robustness and reliability of our algorithm and model both in terms of validation metrics and in terms of a comparison with a 3D visualisation of Z-disks obtained using immunofluorescence based confocal imaging.