Deep learning has proven to be one of the most effective methods in analyzing biological images to extract parameters fundamental for studying physiological functions and pathological conditions. In particular, when coupled with time-lapse microscopy (TLM), deep learning proves particularly effective in studying behaviors involving temporal dynamics. However, TLM videos are often affected by experimental noise and setup limitations, which can lead to inaccurate and poorly reproducible results. Taking advantage of the variational and generative capabilities of Variational Autoencoders (VAEs), we propose VAE-MOTION, a deep learning-based model for the analysis of cardiac contractile dynamics. By incorporating a temporal encoder into its architecture, our model allows the restoration of video quality by removing noise or increasing resolution, while simultaneously extracting accurate contraction-related signals from the latent space. The generation of synthetic videos allowed extensive training of VAE-MOTION, which subsequently validated on real videos from two different cardiac tissue models: 2D monolayers and 3D microtissues. VAE-MOTION was compared to two gold-standard methods in extracting contraction parameters relevant to drug efficacy or toxicity studies, demonstrating its potential for analyzing temporal dynamics in a given phenomenon or process.

In vitro modelling of cardiac microtissues via human pluripotent stem cell-derived cardiomyocytes (hiPSCs-CMs) holds great promise in experimental pharmacology, especially to improve drug safety and efficacy evaluation. However, the predictive capability of these models is currently hampered by a series of heterogeneity sources mainly related, but not limited to, their immature phenotype. While increased levels of complexity in human-derived models can favor maturation, differences in experimental approaches introduce variability in electrophysiological responses, which is difficult to mitigate. To tackle these aspects, this work proposes a novel measurement strategy in the dynamic frequency regime to induce and analyze multiple operation modes in hiPSC-CMs contraction during drug evaluation. We designed and validated a system for fully automated analysis of cardiac models in a dynamic regime via optical time-lapse microscopy. We show that repeated video-based measurements of the same biological samples in the presence of physical conditioning allow, with appropriate normalization and adjustment operations, accurate estimates of the responses of interest. The analyzed operation modes provided increased representativeness of video-based contractile parameters and reliable trend estimates of concentration-response curves without requiring a high number of biological replicates. In particular, the findings obtained with four benchmark experimental models, including two-dimensional (2D), two-dimensional aligned (2DA), tridimensional microtissues (3D-MTs), and engineered heart tissues (EHTs) in triple cell-type co-culturing, indicate that the proposed approach is effective in drug-response evaluation of four pharmacological compounds with known expected effects and a reference solvent, and offers a promising foundation for the development of high-throughput and robust tools for pharmacological screening.

Correction to: Communications Engineeringhttps://doi.org/10.1038/s44172-026-00594-8, published online 11 February 2026 In the version of the article initially published, Or Perlman was missing from the author list and has now been added. Additionally, Alessandro Rizzo was originally listed with two affiliations, but this has now been corrected to a single affiliation. These corrections have been made to the HTML and PDF versions of the article.

Dielectric spectroscopy is a label-free, non-contact, real-time, multi-layer sensing technology, and has been used for identification and quantification of many biological materials. A combination of such sensing features is in demand for monitoring of organ-on-chip systems; however available sensing technologies have yet to address this need. In this work, we explore the possibility of leveraging the inherent features of dielectric spectroscopy for the application in organ-on-chip systems, by investigating three key technological developments using open-ended coaxial probes. Firstly, biocompatible non-contact sensing capabilities are proved by showing similar sensing performance of Parylene C-coated probes and uncoated probes. Secondly, a setup and methodology are developed for highly accurate and non-destructive height positioning of the probe to allow for precise extraction of intermediate sample layers. Finally, non-contact multi-layer sensing performance of the presented technology is successfully demonstrated by means of a biological phantom in a three-layered system. With further integration, dielectric spectroscopy can potentially become a cornerstone sensing technique for organ-on-chip by enabling real-time non-contact tracking of various tissue contents and properties.

We present the fluidic and electrical packaging of a novel silicon-based trans-epithelial electrical resistance (TEER) sensor chip designed for a modular and standardized organ-on-chip (OoC) platform. The package comprises three key components: the housing of the TEER chip, microfluidic routing for seamless integration with the platform, and electrical connections to a platform-integrated potentiostat. This modular solution enables continuous impedance measurements while maintaining unobstructed optical access to the tissue culture region. Experiments confirmed leak-free fluid flow across the stacked microfluidic channels and stable sensitivity of TiN electrodes to PBS. The TEER module retains optical transparency, bi-ocompatibility, and industrial scalability, supporting advanced in situ tissue barrier assessments in standardized OoC systems.

In this work, we present a numerical testbench, realized in a circuit simulation environment, enabling a priori uncertainty evaluation of dielectric spectroscopy in the application field of organs-on-chip. This testbench evaluates the impact of noise, ambient temperature variation and impurity of liquid standards on the uncertainty of dielectric spectroscopy measurements. Moreover, the proposed approach allows to account for the impact on measurement sensitivity of system parameters such as probe dimensions and probe coatings. The estimated uncertainty contributions for the considered effects are compared and benchmarked experimentally. Finally, the testbench is employed to project the dielectric spectroscopy accuracy on a relevant biological application, namely monitoring the growth of a 7 μm-thick kidney cell monolayer.

Engineered heart tissues (EHTs) have shown great potential in recapitulating tissue organization, functions, and cell-cell interactions of the human heart in vitro. Currently, multiple EHT platforms are used by both industry and academia for different applications, such as drug discovery, disease modelling, and fundamental research. The tissues’ contractile force, one of the main hallmarks of tissue function and maturation level of cardiomyocytes, can be read out from EHT platforms by optically tracking the movement of elastic pillars induced by the contractile tissues. However, existing optical tracking algorithms which focus on calculating the contractile force are customized and platform-specific, often not available to the broad research community, and thus hamper head-to-head comparison of the model output. Therefore, there is the need for robust, standardized and platform-independent software for tissues’ force assessment. To meet this need, we developed ForceTracker: a standalone and computationally efficient software for analyzing contractile properties of tissues in different EHT platforms. The software uses a shape-detection algorithm to single out and track the movement of pillars’ tips for the most common shapes of EHT platforms. In this way, we can obtain information about tissues’ contractile performance. ForceTracker is coded in Python and uses a multi-threading approach for time-efficient analysis of large data sets in multiple formats. The software efficiency to analyze circular and rectangular pillar shapes is successfully tested by analyzing different format videos from two EHT platforms, developed by different research groups. We demonstrate robust and reproducible performance of the software in the analysis of tissues over time and in various conditions. ForceTracker’s detection and tracking shows low sensitivity to common incidental defects, such as alteration of tissue shape or air bubbles. Detection accuracy is determined via comparison with manual measurements using the software ImageJ. We developed ForceTracker as a tool for standardized analysis of contractile performance in EHT platforms to facilitate research on disease modeling and drug discovery in academia and industry.

It is estimated that 99 % of the world population is exposed to air pollution above air quality guidelines and this is responsible for 6.7 million premature deaths annually. Lung and skin are the first organs exposed to air pollution, and this is associated with carcinogenesis, inflammation and atopic disease. Proposed mechanisms of adverse health effects in lung and skin include oxidative stress, inflammation, and loss of epithelial barrier integrity. Most knowledge has been gained using simple 2D or more complex culture models, however these cultures have important limitations, such as a lack of perfusion and stretching and lack of cell-cell crosstalk. Organ-on-chip (OoC) technology may be used to overcome limitations of the in vitro models currently used in air pollution research and opens possibilities for studying the pathways underlying adverse health effects of air pollution on immune-mediated diseases of the lung and skin using more physiologically relevant exposure experiments. In this review we discuss currently used in vitro models to study the effect of air pollution on epithelial barrier integrity and development of immune-mediated diseases and identify gaps in current knowledge on adverse health effects of air pollution. We then focus on how OoC technology can enhance mechanistic studies of the skin and lung's response to air pollution.

We present a novel silicon-based organ-on-chip (OoC) device featuring integrated microelectrodes to assess barrier function in biological tissue co-cultures. The microfluidic device consists of two vertically-stacked microchannels separated by a submicron-thin, microporous silicon nitride membrane, enabling in vivo-like proximity for co-cultured tissues. The integrated four-probe electrode geometry on slanted microchannel sidewalls ensures unobstructed optical access to the membrane and consistent measurement repeatability. Experimental validation through electrical impedance spectroscopy supported the device's sensitivity to sodium chloride concentration. Fabricated through a scalable, wafer-scale batch process, the device additionally demonstrated biocompatibility and optical transparency, representing a significant advancement for in situ tissue barrier assessments.

Engineered heart tissues (EHTs) formed around flexible pillars are used to measure the contraction force of myocytes. When based on cardiac cells derived from human induced pluripotent stem cells (hiPSCs), EHTs capture human cardiac physiology and drug responses in vitro. However, variability in contractile function often arises due to variation in tissue positioning on the pillar. Here, novel tapered pillars are introduced to achieve spatial confinement of tissues in EHT devices. The devices are fabricated by moulding polydimethylsiloxane (PDMS) into micromachined tapered cavities of a silicon substrate. The symmetrically-tapered geometry, with the minimum cross-section at the pillar mid-height, restricts tissue movement outside of the indented area. This increases sensitivity and accuracy of tissue contractile readout, providing high reproducibility with reduced variability between data points. Design and stiffness of tapered pillars are investigated to determine the optimal mechanical environment, obtain accurate contractile measurements, and achieve long-term culture of EHTs. Results show that tapered pillars provide superior confinement efficiency (over 90%) compared to straight pillars (30%), with tissue confinement directly correlated to pillar geometry rather than stiffness. The optimized precision in force readouts and long-term tissue studies enables higher sensitivity in the detection of contractile responses to drugs or diseases.

The mechanisms governing the onset and eventual progression of several neurodegenerative disorders remain poorly understood or even undiscovered. This lack of pathophysiological insight can be partly attributed to reliance on inaccurate in vitro models. Notwithstanding research efforts towards recapitulating brain functions on flat devices, mimicking the brain's three-dimensional (3D) architecture in vitro remains a prime target, as 3D models more closely resemble the functional behavior and dynamic responses of in vivo organs. In this work, we present a novel, wafer-scale approach for microfabrication of soft and transparent 3D microelectrode arrays (MEAs) for in vitro electrical recording and optical inspection of electrogenic cell cultures. The proposed 3D MEAs entail 90μ m -high polydimethylsiloxane-based micro-pyramids featuring multiple, electrically-distinct and vertically-stacked titanium nitride electrodes on their slanted facets. Our innovative 3D MEAs will facilitate the development of physiologically-accurate brain-on-a-chip models capable of monitoring 3D electrical communication in neuronal networks while allowing their simultaneous optical characterization.

Real-time pH and oxygen concentration sensing is critical for monitoring tissue damage and organ health; however, there is no report to date in such context of a single device that can simultaneously detect both pH and oxygen changes. This paper presents the development of a single optical sensor device that can simultaneously and reversibly respond to changes in both pH and oxygen concentration. The proposed optical sensor integrates both pH- and oxygen-sensitive probes, and is optimized to achieve minimal cross-sensitivity during simultaneous measurements. This approach enables high accuracy in early detection of chemical correlates of tissue or organ damage, improving screening and efficacy in organ transplants.

Skeletal muscle spatial analyses have revealed unexpected regionalized gene expression patterns challenging the understanding of muscle as a homogeneous tissue. Here, we present a protocol for the spatial analysis of transcript and protein levels in murine skeletal muscle. We describe steps for tibialis anterior dissection, formaldehyde fixation, tissue chopper cutting, and hybridization chain reaction (HCR) detection and amplification. We then detail procedures for immunostaining, tissue clearing, and imaging. This protocol is easily adaptable to other tissues.