Biomedical Imaging Research Articles

A foundation model unlocks unified biomedical image analysis

A foundation model unlocks unified biomedical image analysis

Dual-Locked Chemiluminescent Probe Enables Precise Imaging and Timely Diagnosis of Colitis via Chymotrypsin/Vanin-1 Cascade Activation.

The development of precise diagnosis and the discovery of individualized drugs go together to provide effective therapy against inflammatory bowel disease (IBD). The exploitation of the unique imaging advantages of chemiluminescent probes represents a pivotal strategy for achieving this goal. Nevertheless, the dual-locked strategy, which is believed to enhance precision, is rarely employed in the design of chemiluminescent probes. A novel dual-locked chemiluminescent probe, BPan-CL, was designed based on IBD candidate biomarkers chymotrypsin (CHT) and vanin-1. BPan-CL exhibited specific reactivity and chemiluminescence response when subjected to simultaneous stimulation of CHT and vanin-1, with a signal-to-noise ratio superior to that of the fluorescent probe with the same dual-locked mode. In both live cell and IBD mice imaging, BPan-CL demonstrated superior sensitivity compared to its single-locked counterpart, Pan-CL. In contrast to Pan-CL, BPan-CL was able to more accurately identify IBD and healthy mice by in vivo imaging and allowed for early prediction of IBD using a noninvasive fecal test. BPan-CL has identified CHT and vanin-1 as valuable combinatorial biomarkers for accurate and early IBD diagnosis. This strategy has significant potential for use in biomedical imaging and future individualized therapies.

Metasurface-Based Image Classification Using Diffractive Deep Neural Network

The computer-assisted inverse design of photonic computing, especially by leveraging artificial intelligence algorithms, offers great convenience to accelerate the speed of development and improve calculation accuracy. However, traditional thickness-based modulation methods are hindered by large volume and difficult fabrication process, making it hard to meet the data-driven requirements of flexible light modulation. Here, we propose a diffractive deep neural network (D2NN) framework based on a three-layer all-dielectric phased transmitarray as hidden layers, which can perform the classification of handwritten digits. By tailoring the radius of a silicon nanodisk of a meta-atom, the metasurface can realize the phase profile calculated by D2NN and maintain a relative high transmittance of 0.9 at a wavelength of 600 nm. The designed image classifier consists of three layers of phase-only metasurfaces, each of which contains 1024 units, mimicking a fully connected neural network through the diffraction of light fields. The classification task of handwriting digits from the ‘0’ to ‘5’ dataset is verified, with an accuracy of over 90% on the blind test dataset, as well as demonstrated by the full-wave simulation. Furthermore, the performance of the more complex animal image classification task is also validated by increasing the number of neurons to enhance the connectivity of the neural network. This study may provide a possible solution for practical applications such as biomedical detection, image processing, and machine vision based on all-optical computing.

Motion-induced phase shift for dynamic structured light measurement.

Structured light 3D shape measurement is extensively utilized in semiconductor inspection, smart manufacturing, and biomedical imaging due to its rapid measurement speed, high precision, and versatile applicability to different objects. However, the traditional implementations of this method often require that the object remains static while recording the phase-shifting structured light images, which limits the adaptability of dynamic measurement. Here, we propose a dynamic 3D shape measurement using structured light based on a motion-induced phase shift (MIPS). As the object moves, the surface features distort the fringe pattern, resulting in a phase-shifting effect. By employing the MIPS method, we can determine the phase even in the situations where the knowledge of phase-shifting conditions is not accurate. This enables the acquisition of the 3D topography of the object surface with a high level of precision. Experimental results demonstrate that the MIPS method can accurately measure the 3D shape of objects moving as fast as 100 mm/s, with a relative discrepancy of less than 0.23%.

Antithermal-Quenching Near-Infrared Emitting Lu2SrAl4SiO12:Cr3+ Garnet-Type Phosphor for High-Resolution Nonvisual Imaging.

Near-infrared (NIR) light allows fast and nondestructive detection with deep penetration into biological tissues and is widely used in food inspection, biomedical imaging, night vision security, and other fields. Cr3+-doped NIR first region (NIR-I) phosphors have many interesting features that have attracted a lot of attention recently. However, practical issues, such as low photoluminescence quantum efficiency and poor thermal stability, need to be addressed. We synthesized Cr3+-doped Lu2SrAl4SiO12 garnet-type phosphors using a high-temperature solid-state reaction method. Under blue-light (@ 430 nm) excitation, the phosphors exhibited broadband NIR-I emission in the range of 600-1000 nm, with an emission peak at 710 nm. This system is unique, as the emission had multipeak sharp-line superposition and the Cr3+ ions were situated in a relatively strong crystal field environment, as opposed to a weak crystal field environment for other matrix. The optimal doping content of Cr3+ ions was 2 mol %, and its internal quantum efficiency was ∼76.8%. Surprisingly, this NIR phosphor showed an antithermal quenching effect, and the integrated luminescence intensity of NIR-I emission measured at 573 K was 206.3% of that measured at 303 K. We found that the antithermal quenching of NIR-I luminescence was caused by the extremely low thermal expansion coefficient and rigid structure of the Lu2SrAl4SiO12 matrix in the temperature range, as well as the weakening of electron-phonon coupling with increasing temperature. The optimized phosphor was packaged with a blue chip into a NIR phosphor-converted light-emitting diode device. The light source device showed an output power of 119.02 mW and an electro-optical conversion efficiency of 11.02% under a driving current of 300 mA. The application potential of this NIR phosphor was demonstrated in the field of high-resolution nondestructive imaging.

AI in Biomedical Imaging and Diagnostics

Advances in artificial intelligence (AI) and synthetic biology have profoundly influenced biomedical research, creating transformative opportunities in imaging, diagnostics, and therapeutic engineering. In biomedical imaging, AI-driven algorithms enhance precision and accuracy, enabling automated analysis of complex datasets, real-time imaging insights, and identification of disease biomarkers. Meanwhile, synthetic biology redefines cellular engineering, particularly in T-cell research, by enabling customized functionalities, such as precision-targeted antigen recognition and tunable immune responses. The integration of AI into T-cell engineering amplifies these capabilities, facilitating the design and optimization of synthetic circuits, predictive modeling of cellular behaviors, and dynamic monitoring of therapeutic outcomes. This interdisciplinary approach is revolutionizing diagnostics and immunotherapy by streamlining the identification of disease-specific markers, improving diagnostic accuracy, and enabling real-time modulation of T-cell functionality within the tumor microenvironment. By combining AI-powered insights with synthetic biology's ability to engineer living systems, this research aims to address critical challenges in disease treatment, including tumor heterogeneity and immune evasion. This work explores the synergistic application of AI and synthetic biology in biomedical imaging and T-cell engineering, highlighting state-of-the-art technologies, their therapeutic potential, and the future landscape of personalized medicine.

Spontaneous breaking of symmetry in overlapping cell instance segmentation using diffusion models

Abstract Instance segmentation is the task of assigning unique identifiers to individual objects in images. Solving this task requires breaking the inherent symmetry that semantically similar objects must result in distinct outputs. Deep learning algorithms bypass this break-of-symmetry by training specialized predictors or by utilizing intermediate label representations. However, many of these approaches break down when faced with overlapping labels that are ubiquitous in biomedical imaging, for instance for segmenting cell layers. Here, we discuss the reason for this failure and offer a novel approach for instance segmentation based on diffusion models that breaks this symmetry spontaneously. Our method outputs pixel-level instance segmentations matching the performance of models such as cellpose on the cellpose fluorescent cell dataset while also permitting overlapping labels.

Ultra‐Sensitive, Self‐powered, CMOS‐Compatible Near‐Infrared Photodetectors for Wide‐Ranging Applications

AbstractSelf‐powered near‐infrared (NIR) photodetectors are essential for surveillance systems, sensing in IoT electronics, facial recognition, health monitoring, optical communication networks, night vision, and biomedical imaging. However, silicon commercial detectors need external power to operate and cooling to suppress large dark currents. This work demonstrates a new class of CMOS‐compatible self‐powered NIR photodetector based on ferroelectric 5‐nm thick ZrO2 films which do not require cooling and therefore have two key advantages over Si, and at the same time have comparable performance metrics. At room‐temperature, under 940 nm wavelength illumination (1.4 mW cm−2 power density, 10 Hz repetition rate), and without any power applied, fast rise and fall times of ≈2 and 4 µs, respectively, are achieved in Al/Si/SiOx/ZrO2/ITO devices, along with responsivity, detectivity and sensitivity values of up to ≈3.4 A W−1, 1.2 × 1010 Jones and 4.2 × 103, respectively, far exceeding all other emerging self‐powered systems. Furthermore, dual‐band NIR detection is shown for different NIR wavelengths, proof‐of‐concept feasibility being demonstrated for the smart identification of NIR targets. Therefore, it is demonstrated, for the first time, that coupling together the pyroelectric effect, the photovoltaic effect, and the ferroelectric effect is a novel method to significantly enhance the performance of CMOS‐compatible ZrO2‐based self‐powered photodetectors in the NIR region.

Advances in Lanthanide-Based NIR-IIb Probes for In Vivo Biomedical Imaging.

The past decades have witnessed the significant development and practical interest of in vivo biomedical imaging technologies and optical materials in the second-near infrared (NIR-II, 1000-1700 nm) window. Imaging with the extended emission wavelength toward the long-wavelength end (NIR-IIb, 1500-1700 nm) further offers micrometer imaging resolution and centimeter tissue penetration depth by taking advantage of the much-reduced photon scattering and near-zero tissue autofluorescence background, which have become a very hot research area. This review focuses on the recent advances in the development of lanthanide-based NIR-IIb probes for in vivo biomedical applications. The progress including ratiometric imaging, multiplexed imaging for wide-field and microscopy, lifetime multiplexing and sensing, persistent luminescence, and multimodal imaging is summarized. Challenges and future directions concerning the investigation of the photophysical and photochemical properties of NIR-IIb probes, the selection of near-infrared cameras as well as the potential extension of the NIR-IIb imaging sub-window are pointed out. This review will inspire readers who have a strong interest in developing optical imaging technology and long-wavelength fluorescence probes for high-contrast in vivo biomedical applications.

Self-supervised learning for chest computed tomography: training strategies and effect on downstream applications.

Self-supervised pre-training can reduce the amount of labeled training data needed by pre-learning fundamental visual characteristics of the medical imaging data. We investigate several self-supervised training strategies for chest computed tomography exams and their effects on downstream applications. We benchmark five well-known self-supervision strategies (masked image region prediction, next slice prediction, rotation prediction, flip prediction, and denoising) on 15M chest computed tomography (CT) slices collected from four sites of the Mayo Clinic enterprise, United States. These models were evaluated for two downstream tasks on public datasets: pulmonary embolism (PE) detection (classification) and lung nodule segmentation. Image embeddings generated by these models were also evaluated for prediction of patient age, race, and gender to study inherent biases in models' understanding of chest CT exams. The use of pre-training weights especially masked region prediction-based weights, improved performance, and reduced computational effort needed for downstream tasks compared with task-specific state-of-the-art (SOTA) models. Performance improvement for PE detection was observed for training dataset sizes as large as with a maximum gain of 5% over SOTA. The segmentation model initialized with pre-training weights learned twice as fast as the randomly initialized model. While gender and age predictors built using self-supervised training weights showed no performance improvement over randomly initialized predictors, the race predictor experienced a 10% performance boost when using self-supervised training weights. We released self-supervised models and weights under an open-source academic license. These models can then be fine-tuned with limited task-specific annotated data for a variety of downstream imaging tasks, thus accelerating research in biomedical imaging informatics.

Deep learning for 3D vascular segmentation in hierarchical phase contrast tomography: a case study on kidney

Automated blood vessel segmentation is critical for biomedical image analysis, as vessel morphology changes are associated with numerous pathologies. Still, precise segmentation is difficult due to the complexity of vascular structures, anatomical variations across patients, the scarcity of annotated public datasets, and the quality of images. Our goal is to provide a foundation on the topic and identify a robust baseline model for application to vascular segmentation using a new imaging modality, Hierarchical Phase-Contrast Tomography (HiP-CT). We begin with an extensive review of current machine-learning approaches for vascular segmentation across various organs. Our work introduces a meticulously curated training dataset, verified by double annotators, consisting of vascular data from three kidneys imaged using HiP-CT as part of the Human Organ Atlas Project. HiP-CT pioneered at the European Synchrotron Radiation Facility in 2020, revolutionizes 3D organ imaging by offering a resolution of around 20 μm/voxel and enabling highly detailed localised zooms up to 1–2 μm/voxel without physical sectioning. We leverage the nnU-Net framework to evaluate model performance on this high-resolution dataset, using both known and novel samples, and implementing metrics tailored for vascular structures. Our comprehensive review and empirical analysis on HiP-CT data sets a new standard for evaluating machine learning models in high-resolution organ imaging. Our three experiments yielded Dice similarity coefficient (DSC) scores of 0.9523, 0.9410, and 0.8585, respectively. Nevertheless, DSC primarily assesses voxel-to-voxel concordance, overlooking several crucial characteristics of the vessels and should not be the sole metric for deciding the performance of vascular segmentation. Our results show that while segmentations yielded reasonably high scores-such as centerline DSC ranging from 0.82 to 0.88, certain errors persisted. Specifically, large vessels that collapsed due to the lack of hydrostatic pressure (HiP-CT is an ex vivo technique) were segmented poorly. Moreover, decreased connectivity in finer vessels and higher segmentation errors at vessel boundaries were observed. Such errors, particularly in significant vessels, obstruct the understanding of the structures by interrupting vascular tree connectivity. Our study establishes the benchmark across various evaluation metrics, for vascular segmentation of HiP-CT imaging data, an imaging technology that has the potential to substantively shift our understanding of human vascular networks.

Blazing Carbon Dots: Unfolding its Luminescence Mechanism to Photoinduced Biomedical Applications.

Carbon dots (CDs) are carbon-based nanomaterials that have garnered immense interest due to their exceptional photophysical and optoelectronic properties. They have been employed extensively for biomedical imaging and phototherapy due to their superb water dispersibility, low toxicity, outstanding biocompatibility, and exceptional tissue permeability. This review summarizes the luminescence mechanism of CDs. The modification in CDs via various doping routes is comprehensively reviewed, and the effect of such alterations on their photophysical properties, such as photoluminescence (PL), absorbance, and reactive oxygen species generation ability, is also highlighted. This review also aims to summarize the role of CDs in cellular imaging and fluorescence lifetime imaging for cellular metabolism. Subsequently, recent advancements and the future prospects of CDs as nanotheranostic agents have been discussed. Herein, we have discussed the role of CDs in photothermal, photodynamic, and synergistic therapy of anticancer, antiviral, and antibacterial applications. The overall summary of the review highlights the future prospects of CD-based research in bioimaging and biomedicine.

Ultrafast Polarization‐Maintaining Fiber Lasers: Design, Fabrication, Performance, and Applications

AbstractUltrafast polarization‐maintaining fiber lasers (UPMFLs), with superior optical performance and high immunity to environmental disturbances, are highly preferable in a variety of industrial and scientific applications such as high‐precision micromachining and biomedical imaging. Especially, the utilization of PM fibers endows the laser intrinsic stability, thereby enabling the construction of robust and low‐noise optical frequency comb systems. To meet more demanding application challenges, continuous efforts have been invested in the design and fabrication of UPMFLs, aiming to reach unprecedented levels of various pulse parameters, that is, to achieve shorter pulse duration, higher or lower repetition rate, and higher pulse energy. This review presents a detailed overview of different passive mode‐locking techniques for pulsed operation and the most significant achievements in UPMFLs. Representative advances at 1.0, 1.55, and 2.0 µm spectral regions are presented and summarized. The state‐of‐the‐art lasing performance is application‐oriented, and conversely, optical improvements in all‐PM pulsed lasers promote emerging applications, which are also discussed and analyzed. How to overcome the bottlenecks of UPMFLs in terms of pulse duration, repetition rate, emission wavelength, and pulse energy to make them powerful tools for physical, medical, and biological applications remains challenging in the future.

Cutting-Edge Machine Learning in Biomedical Image Analysis: Editorial for Bioengineering Special Issue: “Recent Advance of Machine Learning in Biomedical Image Analysis”

Biomedical image analysis plays a critical role in the healthcare system [...]

Acoustic field visualization and source localization via physics-informed learning of sparse data with adaptive sampling

Traditional methods for acoustic field visualization require considerable effort for capturing large amounts of acoustic data to achieve a high resolution field map, highly limiting their widespread use. In this study, we propose an approach for acoustic field visualization based on physics-informed neural networks (PINNs) by using a small amount of data, subsequently realizing accurate acoustic source localization. First, we present a PINN model integrated with an acoustic Helmholtz equation and adaptive sampling, the performance of which is testified via numerical simulations. The “no mesh” character of PINN enables achieving high resolution acoustic field visualization without requiring the capture of numerous data in advance. Furthermore, we experimentally validate the performance of the proposed method, which demonstrates that the acoustic sources can be precisely localized with sparse field data acquisition within a small area. This work would find potential applications in various acoustics, such as acoustic communication, biomedical imaging, and virtual reality.

Metal-Phenolic Networks: A Promising Frontier in Cancer Theranostics.

The burgeoning field of cancer theranostics has been significantly advanced by the development of Metal-Phenolic Networks (MPNs), a new class of supramolecular architectures that integrate the advantages of metals and polyphenols. This review focuses on MPNs and their promising applications in cancer theranostics. Through a systematic literature search spanning from 2010 to 2023 in databases including PubMed, Scopus, and Web of Science. The period of search was justified by the rapid evolution of nanomaterials in cancer therapy, with MPNs emerging as a significant player in biomedical applications within the specified timeframe. This review discusses the classification and structure of polyphenolic compounds, as well as their mechanisms of action in cancer treatment. The applications of MPNs in chemotherapy drug delivery, photothermal therapy, chemodynamic therapy, biomedical imaging, and synergistic therapy are especially detailed. The authors emphasize the significance of MPNs in cancer nanomedicine and look forward to their future development directions.

Design of eye phantom for optical coherence tomography angiography study.

Tissue phantoms play an important role in validating biomedical imaging techniques. For optical imaging systems such as optical coherence tomography, creating layers of phantoms with different refractive indices is important. This paper explored various phantom materials to create eye-mimicking phantoms. The phantom also included microfluidic channels to mimic vasculature. To verify the optical properties of the created phantom, the transmission and reflection spectroscopy was performed using a photoluminescent spectrometer studying the complete visible and near infrared spectrum. We created a 3D model to house multiple layers that mimicked the retina at one end and the contact lens at the other, representing the anterior segment of the eye. A realistic retinal layer-mimicking, which, among other things, may pave the way for new combined imaging modalities.

Leveraging deep learning for identification and segmentation of “CAF-1/p60-positive” nuclei in oral squamous cell carcinoma tissue samples

In the current study, we introduced a unique method for identifying and segmenting oral squamous cell carcinoma (OSCC) nuclei, concentrating on those predicted to have significant CAF-1/p60 protein expression. Our suggested model uses the StarDist architecture, a deep-learning framework designed for biomedical image segmentation tasks. The training dataset comprises painstakingly annotated masks created from tissue sections previously stained with hematoxylin and eosin (H&E) and then restained with immunohistochemistry (IHC) for p60 protein. Our algorithm uses subtle morphological and colorimetric H&E cellular characteristics to predict CAF-1/p60 IHC expression in OSCC nuclei. The StarDist-based architecture performs exceptionally well in localizing and segmenting H&E nuclei, previously identified by IHC-based ground truth. In summary, our innovative approach harnesses deep learning and multimodal information to advance the automated analysis of OSCC nuclei exhibiting specific protein expression patterns. This methodology holds promise for expediting accurate pathological assessment and gaining deeper insights into the role of CAF-1/p60 protein within the context of oral cancer progression.

Enhancing the Security and Efficiency of Biomedical Image Encryption through a Novel Hyper-Chaotic Logistic Map

Enhancing the Security and Efficiency of Biomedical Image Encryption through a Novel Hyper-Chaotic Logistic Map

224-fs soliton pulses generation at 1μm from ytterbium-doped fiber laser with CoTe2 nanosheets as an ultrafast modulator

Transition metal ditellurides (TMDTs) have numerous attractive properties, making them suitable for a wide range of applications. In this study, cobalt ditelluride (CoTe2) nanosheets, a promising TMDT for photonic applications, were prepared using an ultrasound-enhanced liquid phase exfoliation (LPE) method. A novel saturable absorber (SA) employing CoTe2 nanosheets was then fabricated by optically depositing them on microfiber. The nonlinear optical modulation properties of the CoTe2 SA were investigated. A high-performance 1 μm ultrafast fiber laser was demonstrated by incorporating newly developed CoTe2 nanosheets-based SA in a ring cavity ytterbium-doped fiber laser (YDFL). The dynamical behaviour of the proposed passively mode-locked YDFL in response to variations in pump optical power was investigated. The findings reveal that the device achieved a modulation depth of 2.5 %, and saturation light intensity of 30.6 MW/cm2. Moreover, a stable and robust mode-locked soliton optical pulse sequence with a fundamental repetition frequency of 3.089 MHz, and a pulse duration of 224 fs was generated at 1032 nm. The proposed YDFL, being all-fiber, compact, and cost-effective, is set to find extensive applications in various domains, including optical fiber communication, sensing, and biomedical imaging.