New open-source software for subcellular segmentation and analysis of spatiotemporal fluorescence signals using deep learning

SANS-CNN: An automated machine learning technique for spaceflight associated neuro-ocular syndrome with astronaut imaging data

Spaceflight associated neuro-ocular syndrome (SANS) is one of the largest physiologic barriers to spaceflight and requires evaluation and mitigation for future planetary missions. As the spaceflight environment is a clinically limited environment, the purpose of this research is to provide automated, early detection and prognosis of SANS with a machine learning model trained and validated on astronaut SANS optical coherence tomography (OCT) images. In this study, we present a lightweight convolutional neural network (CNN) incorporating an EfficientNet encoder for detecting SANS from OCT images titled "SANS-CNN." We used 6303 OCT B-scan images for training/validation (80%/20% split) and 945 for testing with a combination of terrestrial images and astronaut SANS images for both testing and validation. SANS-CNN was validated with SANS images labeled by NASA to evaluate accuracy, specificity, and sensitivity. To evaluate real-world outcomes, two state-of-the-art pre-trained architectures were also employed on this dataset. We use GRAD-CAM to visualize activation maps of intermediate layers to test the interpretability of SANS-CNN's prediction. SANS-CNN achieved 84.2% accuracy on the test set with an 85.6% specificity, 82.8% sensitivity, and 84.1% F1-score. Moreover, SANS-CNN outperforms two other state-of-the-art pre-trained architectures, ResNet50-v2 and MobileNet-v2, in accuracy by 21.4% and 13.1%, respectively. We also apply two class-activation map techniques to visualize critical SANS features perceived by the model. SANS-CNN represents a CNN model trained and validated with real astronaut OCT images, enabling fast and efficient prediction of SANS-like conditions for spaceflight missions beyond Earth's orbit in which clinical and computational resources are extremely limited.

Ca2+ dynamics in interstitial cells: foundational mechanisms for the motor patterns in the gastrointestinal tract

The gastrointestinal (GI) tract displays multiple motor patterns that move nutrients and wastes through the body. Smooth muscle cells (SMCs) provide the forces necessary for GI motility, but interstitial cells, electrically coupled to SMCs, tune SMC excitability, transduce inputs from enteric motor neurons, and generate pacemaker activity that underlies major motor patterns, such as peristalsis and segmentation. The interstitial cells regulating SMCs are interstitial cells of Cajal (ICC) and PDGF receptor (PDGFR)α+ cells. Together these cells form the SIP syncytium. ICC and PDGFRα+ cells express signature Ca2+-dependent conductances: ICC express Ca2+-activated Cl- channels, encoded by Ano1, that generate inward current, and PDGFRα+ cells express Ca2+-activated K+ channels, encoded by Kcnn3, that generate outward current. The open probabilities of interstitial cell conductances are controlled by Ca2+ release from the endoplasmic reticulum. The resulting Ca2+ transients occur spontaneously in a stochastic manner. Ca2+ transients in ICC induce spontaneous transient inward currents and spontaneous transient depolarizations (STDs). Neurotransmission increases or decreases Ca2+ transients, and the resulting depolarizing or hyperpolarizing responses conduct to other cells in the SIP syncytium. In pacemaker ICC, STDs activate voltage-dependent Ca2+ influx, which initiates a cluster of Ca2+ transients and sustains activation of ANO1 channels and depolarization during slow waves. Regulation of GI motility has traditionally been described as neurogenic and myogenic. Recent advances in understanding Ca2+ handling mechanisms in interstitial cells and how these mechanisms influence motor patterns of the GI tract suggest that the term "myogenic" should be replaced by the term "SIPgenic," as this review discusses.

Live-cell imaging in the deep learning era

Live imaging is a powerful tool, enabling scientists to observe living organisms in real time. In particular, when combined with fluorescence microscopy, live imaging allows the monitoring of cellular components with high sensitivity and specificity. Yet, due to critical challenges (i.e., drift, phototoxicity, dataset size), implementing live imaging and analyzing the resulting datasets is rarely straightforward. Over the past years, the development of bioimage analysis tools, including deep learning, is changing how we perform live imaging. Here we briefly cover important computational methods aiding live imaging and carrying out key tasks such as drift correction, denoising, super-resolution imaging, artificial labeling, tracking, and time series analysis. We also cover recent advances in self-driving microscopy.

Machine learning (ML)-assisted surface tension and oscillation-induced elastic modulus studies of oxide-coated liquid metal (LM) alloys

Pendant drops of oxide-coated high-surface tension fluids frequently produce perturbed shapes that impede interfacial studies. Eutectic gallium indium or Galinstan are high-surface tension fluids coated with a ∼5 nm gallium oxide (Ga2O3) film and falls under this fluid classification, also known as liquid metals (LMs). The recent emergence of LM-based applications often cannot proceed without analyzing interfacial energetics in different environments. While numerous techniques are available in the literature for interfacial studies- pendant droplet-based analyses are the simplest. However, the perturbed shape of the pendant drops due to the presence of surface oxide has been ignored frequently as a source of error. Also, exploratory investigations of surface oxide leveraging oscillatory pendant droplets have remained untapped. We address both challenges and present two contributing novelties- (a) by utilizing the machine learning (ML) technique, we predict the approximate surface tension value of perturbed pendant droplets, (ii) by leveraging the oscillation-induced bubble tensiometry method, we study the dynamic elastic modulus of the oxide-coated LM droplets. We have created our dataset from LM's pendant drop shape parameters and trained different models for comparison. We have achieved >99% accuracy with all models and added versatility to work with other fluids. The best-performing model was leveraged further to predict the approximate values of the nonaxisymmetric LM droplets. Then, we analyzed LM's elastic and viscous moduli in air, harnessing oscillation-induced pendant droplets, which provides complementary opportunities for interfacial studies alternative to expensive rheometers. We believe it will enable more fundamental studies of the oxide layer on LM, leveraging both symmetric and perturbed droplets. Our study broadens the materials science horizon, where researchers from ML and artificial intelligence domains can work synergistically to solve more complex problems related to surface science, interfacial studies, and other studies relevant to LM-based systems.

Generative Adversarial Networks in Medicine: Important Considerations for this Emerging Innovation in Artificial Intelligence

The advent of artificial intelligence (AI) and machine learning (ML) has revolutionized the field of medicine. Although highly effective, the rapid expansion of this technology has created some anticipated and unanticipated bioethical considerations. With these powerful applications, there is a necessity for framework regulations to ensure equitable and safe deployment of technology. Generative Adversarial Networks (GANs) are emerging ML techniques that have immense applications in medical imaging due to their ability to produce synthetic medical images and aid in medical AI training. Producing accurate synthetic images with GANs can address current limitations in AI development for medical imaging and overcome current dataset type and size constraints. Offsetting these constraints can dramatically improve the development and implementation of AI medical imaging and restructure the practice of medicine. As observed with its other AI predecessors, considerations must be taken into place to help regulate its development for clinical use. In this paper, we discuss the legal, ethical, and technical challenges for future safe integration of this technology in the healthcare sector.

Algorithm for biological second messenger analysis with dynamic regions of interest

Physiological function is regulated through cellular communication that is facilitated by multiple signaling molecules such as second messengers. Analysis of signal dynamics obtained from cell and tissue imaging is difficult because of intricate spatially and temporally distinct signals. Signal analysis tools based on static region of interest analysis may under- or overestimate signals in relation to region of interest size and location. Therefore, we developed an algorithm for biological signal detection and analysis based on dynamic regions of interest, where time-dependent polygonal regions of interest are automatically assigned to the changing perimeter of detected and segmented signals. This approach allows signal profiles to be rigorously and precisely tracked over time, eliminating the signal distortion observed with static methods. Integration of our approach with state-of-the-art image processing and particle tracking pipelines enabled the isolation of dynamic cellular signaling events and characterization of biological signaling patterns with distinct combinations of parameters including amplitude, duration, and spatial spread. Our algorithm was validated using synthetically generated datasets and compared with other available methods. Application of the algorithm to volumetric time-lapse hyperspectral images of cyclic adenosine monophosphate measurements in rat microvascular endothelial cells revealed distinct signal heterogeneity with respect to cell depth, confirming the utility of our approach for analysis of 5-dimensional data. In human tibial arteries, our approach allowed the identification of distinct calcium signal patterns associated with atherosclerosis. Our algorithm for automated detection and analysis of second messenger signals enables the decoding of signaling patterns in diverse tissues and identification of pathologic cellular responses.

Software for segmenting and quantifying calcium signals using multi-scale generative adversarial networks

Cellular calcium fluorescence imaging utilized to study cellular behaviors typically results in large datasets and a profound need for standardized and accurate analysis methods. Here, we describe open-source software (4SM) to overcome these limitations using an automated machine learning pipeline for subcellular calcium signal segmentation of spatiotemporal maps. The primary use of 4SM is to analyze spatiotemporal maps of calcium activities within cells or across multiple cells. For complete details on the use and execution of this protocol, please refer to Kamran et al. (2022).¹.