Whole-cell segmentation of tissue images with human-level performance using large-scale data annotation and deep learning.

A principal challenge in the analysis of tissue imaging data is cell segmentation-the task of identifying the precise boundary of every cell in an image. To address this problem we constructed TissueNet, a dataset for training segmentation models that contains more than 1 million manually labeled cells, an order of magnitude more than all previously published segmentation training datasets. We used TissueNet to train Mesmer, a deep-learning-enabled segmentation algorithm. We demonstrated that Mesmer is more accurate than previous methods, generalizes to the full diversity of tissue types and imaging platforms in TissueNet, and achieves human-level performance. Mesmer enabled the automated extraction of key cellular features, such as subcellular localization of protein signal, which was challenging with previous approaches. We then adapted Mesmer to harness cell lineage information in highly multiplexed datasets and used this enhanced version to quantify cell morphology changes during human gestation. All code, data and models are released as a community resource.

DeepCell Kiosk: scaling deep learning-enabled cellular image analysis with Kubernetes.

Deep learning is transforming the analysis of biological images, but applying these models to large datasets remains challenging. Here we describe the DeepCell Kiosk, cloud-native software that dynamically scales deep learning workflows to accommodate large imaging datasets. To demonstrate the scalability and affordability of this software, we identified cell nuclei in 106 1-megapixel images in ~5.5 h for ~US$250, with a cost below US$100 achievable depending on cluster configuration. The DeepCell Kiosk can be downloaded at https://github.com/vanvalenlab/kiosk-console ; a persistent deployment is available at https://deepcell.org/ .

Deep learning for cellular image analysis.

Recent advances in computer vision and machine learning underpin a collection of algorithms with an impressive ability to decipher the content of images. These deep learning algorithms are being applied to biological images and are transforming the analysis and interpretation of imaging data. These advances are positioned to render difficult analyses routine and to enable researchers to carry out new, previously impossible experiments. Here we review the intersection between deep learning and cellular image analysis and provide an overview of both the mathematical mechanics and the programming frameworks of deep learning that are pertinent to life scientists. We survey the field's progress in four key applications: image classification, image segmentation, object tracking, and augmented microscopy. Last, we relay our labs' experience with three key aspects of implementing deep learning in the laboratory: annotating training data, selecting and training a range of neural network architectures, and deploying solutions. We also highlight existing datasets and implementations for each surveyed application.

The interphase interval within a bipolar nanosecond electric pulse modulates bipolar cancellation.

Nanosecond electric pulse (nsEP) exposure generates an array of physiological effects. The extent of these effects is impacted by whether the nsEP is a unipolar (UP) or bipolar (BP) exposure. A 600 ns pulse can generate 71% more YO-PRO-1 uptake compared to a 600 ns + 600 ns pulse exposure. This observation is termed "bipolar cancellation" (BPC) because despite the BP nsEP consisting of an additional 600 ns pulse, it generates reduced membrane perturbation. BPC is achieved by varying pulse amplitudes, and symmetrical and asymmetric pulse widths. The effect appears to reverse by increasing the interphase interval between symmetric BP pulses, suggesting membrane recovery is a BPC factor. To date, the impact of the interphase interval between asymmetrical BP and other BPC-inducing symmetrical BP nsEPs has not been fully explored. Additionally, interpulse intervals beyond 50 µs have not been explored to understand the impact of time between the BP nsEP phases. Here, we surveyed different interphase intervals among symmetrical and asymmetrical BP nsEPs to monitor their impact on BPC of YO-PRO-1 uptake. We identified that a 10 microsecond (ms) interphase interval within a symmetrical 600 ns + 600 ns, and 900 ns + 900 ns pulse can resolve BPC. Furthermore, the interphase interval to resolve asymmetric BPC from a 300 ns + 900 ns pulse versus 600 ns pulse exposure is greater (<10 ms) compared to symmetrical BP nsEPs. From these findings, we extended on our conceptual model that BPC is balanced by localized charging and discharging events across the membrane. Bioelectromagnetics. 39:441-450, 2018. Published 2018. This article is a U.S. Government work and is in the public domain in the USA.

Asymmetrical bipolar nanosecond electric pulse widths modify bipolar cancellation.

A bipolar (BP) nanosecond electric pulse (nsEP) exposure generates reduced calcium influx compared to a unipolar (UP) nsEP. This attenuated physiological response from a BP nsEP exposure is termed "bipolar cancellation" (BPC). The predominant BP nsEP parameters that induce BPC consist of a positive polarity (↑) front pulse followed by the delivery of a negative polarity (↓) back pulse of equal voltage and width; thereby the duration is twice a UP nsEP exposure. We tested these BPC parameters, and discovered that a BP nsEP with symmetrical pulse widths is not required to generate BPC. For example, our data revealed the physiological response initiated by a ↑900 nsEP exposure can be cancelled by a second pulse that is a third of its duration.  However, we observed a complete loss of BPC from a ↑300 nsEP followed by a ↓900 nsEP exposure. Spatiotemporal analysis revealed these asymmetrical BP nsEP exposures generate distinct local YO-PRO®-1 uptake patterns across the plasma membrane. From these findings, we generated a conceptual model that suggests BPC is a phenomenon balanced by localized charging and discharging events across the membrane.

Probe beam deflection optical imaging of thermal and mechanical phenomena resulting from nanosecond electric pulse (nsEP) exposure in-vitro.

Electric-field induced physical phenomena, such as thermal, mechanical and electrochemical dynamics, may be the driving mechanism behind bioeffects observed in mammalian cells during exposure to nanosecond-duration electric pulses (nsEP) in-vitro. Correlating a driving mechanism to a biological response requires the experimental measurement and quantification of all physical dynamics resulting from the nsEP stimulus. A passive and electromagnetic interference (EMI) immune sensor is required to resolve these dynamics in high strength electric fields. The probe beam deflection technique (PBDT) is a passive and EMI immune optical method for quantifying and imaging refractive index gradients in liquids and gases, both dynamic and static, with nanosecond temporal resolution. In this work, a probe beam deflection imaging system was designed to acquire 2-D time-lapse images of thermal/mechanical dynamics resulting from monopolar and bipolar nsEP stimulus.

Quantifying pulsed electric field-induced membrane nanoporation in single cells.

Plasma membrane disruption can trigger a host of cellular activities. One commonly observed type of disruption is pore formation. Molecular dynamic (MD) simulations of simplified lipid membrane structures predict that controllably disrupting the membrane via nano-scale poration may be possible with nanosecond pulsed electric fields (nsPEF). Until recently, researchers hoping to verify this hypothesis experimentally have been limited to measuring the relatively slow process of fluorescent markers diffusing across the membrane, which is indirect evidence of nanoporation that could be channel-mediated. Leveraging recent advances in nonlinear optical microscopy, we elucidate the role of pulse parameters in nsPEF-induced membrane permeabilization in live cells. Unlike previous techniques, it is able to directly observe loss of membrane order at the onset of the pulse. We also develop a complementary theoretical model that relates increasing membrane permeabilization to membrane pore density. Due to the significantly improved spatial and temporal resolution possible with our imaging method, we are able to directly compare our experimental and theoretical results. Their agreement provides substantial evidence that nanoporation does occur and that its development is dictated by the electric field distribution.

Evaluation of the Genetic Response of U937 and Jurkat Cells to 10-Nanosecond Electrical Pulses (nsEP).

Nanosecond electrical pulse (nsEP) exposure activates signaling pathways, produces oxidative stress, stimulates hormone secretion, causes cell swelling and induces apoptotic and necrotic death. The underlying biophysical connection(s) between these diverse cellular reactions and nsEP has yet to be elucidated. Using global genetic analysis, we evaluated how two commonly studied cell types, U937 and Jurkat, respond to nsEP exposure. We hypothesized that by studying the genetic response of the cells following exposure, we would gain direct insight into the stresses experienced by the cell and in turn better understand the biophysical interaction taking place during the exposure. Using Ingenuity Systems software, we found genes associated with cell growth, movement and development to be significantly up-regulated in both cell types 4 h post exposure to nsEP. In agreement with our hypothesis, we also found that both cell lines exhibit significant biological changes consistent with mechanical stress induction. These results advance nsEP research by providing strong evidence that the interaction of nsEPs with cells involves mechanical stress.

iBET: Immersive visualization of biological electron-transfer dynamics.

Recently, we presented a computational framework named VizBET to simulate and visualize biological electron-transfer (ET) dynamics. The visualization process was encapsulated as a plugin to the Visual Molecular Dynamics (VMD) software. However, the user's ability to understand complex, multidimensional ET pathways was severely limited when visualized in 2D on traditional computer monitors. To provide a more accurate representation with enhanced depth perception, we here present an extension of VizBET named iBET to render the VMD model of ET dynamics in a commodity virtual reality (VR) platform. The paper describes detailed procedures to export VMD models into the Unity game engine and render it in an Oculus Rift head mounted display. With the increasing availability of low-cost VR systems like the Rift and rich programmability of game engines, the iBET framework provides a powerful means to explore and understand not only biological ET processes but also a unique experiential tool for broad scientific communities.

Detecting subtle plasma membrane perturbation in living cells using second harmonic generation imaging.

The requirement of center asymmetry for the creation of second harmonic generation (SHG) signals makes it an attractive technique for visualizing changes in interfacial layers such as the plasma membrane of biological cells. In this article, we explore the use of lipophilic SHG probes to detect minute perturbations in the plasma membrane. Three candidate probes, Di-4-ANEPPDHQ (Di-4), FM4-64, and all-trans-retinol, were evaluated for SHG effectiveness in Jurkat cells. Di-4 proved superior with both strong SHG signal and limited bleaching artifacts. To test whether rapid changes in membrane symmetry could be detected using SHG, we exposed cells to nanosecond-pulsed electric fields, which are believed to cause formation of nanopores in the plasma membrane. Upon nanosecond-pulsed electric fields exposure, we observed an instantaneous drop of ~50% in SHG signal from the anodic pole of the cell. When compared to the simultaneously acquired fluorescence signals, it appears that the signal change was not due to the probe diffusing out of the membrane or changes in membrane potential or fluidity. We hypothesize that this loss in SHG signal is due to disruption in the interfacial nature of the membrane. The results show that SHG imaging has great potential as a tool for measuring rapid and subtle plasma membrane disturbance in living cells.