Text guided 3D medical image segmentation offers a flexible alternative to class based and spatial prompt based models by allowing users to specify regions of interest directly in natural language. This paradigm avoids reliance on predefined label sets, reduces ambiguous outputs, and aligns more naturally with clinical workflows. However, existing text guided frameworks are often computationally expensive, exhibit weak text volume feature alignment, and fail to capture fine anatomical details. We propose ESICA, a lightweight and scalable framework that addresses these challenges through three innovations: (1) a similarity matrix based mask prediction formulation that enhances semantic alignment, (2) an efficient decomposed decoder with adapter modules for accurate volumetric decoding, and (3) a two pass refinement strategy that sharpens boundaries and resolves uncertain regions. To improve training stability and generalization, ESICA adopts a two stage scheme consisting of positive only pretraining followed by balanced fine tuning. On the CVPR BiomedSegFM benchmark spanning five imaging modalities (CT, MRI, PET, ultrasound, and microscopy), ESICA achieves state of the art segmentation accuracy, while the compact ESICA4 Lite variant attains similar segmentation performance with substantially fewer parameters, yielding a superior efficiency accuracy trade off. Our framework advances text guided segmentation toward efficient, scalable, and clinically deployable systems. Code will be made publicly available at https://github.com/mirthAI/ESICA.

With the advancement of high-end optical manufacturing toward higher precision and larger apertures, the quality requirements for optical components have been continuously increasing. As a non-contact and high-precision measurement instrument, interferometers are commonly employed to characterize the surface of optical elements. However, as the aperture size of components continues to grow, stress-induced birefringence in both the system and the sample introduces additional optical path differences, thereby affecting the accuracy of interferometric measurements. At present, polarization errors in optical elements are typically evaluated using dedicated instruments based on the birefringence effect. Nevertheless, due to aperture limitations of such instruments, in-situ measurements of large-aperture optical components remain infeasible. In this paper, we propose a stress birefringence index method employing a liquid crystal variable phase retarder (LCVR), which modulates the phase delay of polarized light and calculates the delay of the sample from the intensity images. This method enables the measurement of stress-induced birefringence in optical components without mechanically rotating polarizing elements, offering the advantage of fast and motionless. Validation experiments were conducted by comparison with the Senarmont method by rotating the analyzer, that demonstrate the reliability of the proposed approach. Finally, the method was applied in a 100 mm Fizeau interferometric system, where a phase delay of 3.4 nm/cm was measured for a collimating element.

This study proposes a method for simultaneously measuring the topological charge (TC) and the power-exponent parameter of power-exponent-phase vortices (PEPVs) using an elliptical Airyprime (EAP) phase mask. It is found that the first main lobe length of the diffraction pattern from EAP is strongly correlated with the power-exponent parameter of detected PEPVs. Both the value and sign of the TC, along with the power-exponent parameter, can be determined from the diffraction pattern without altering the phase mask parameters or the detection position. Experimental results confirmed the reliability of the proposed method, providing an efficient solution for real-time, high-throughput measurement of optical vortex beams.

Spatially-resolved characterization of absorptive defects on high-quality fused silica optical surfaces is demonstrated by combining laser thermal pumping with dynamic micro-interferometric imaging (LTP-DMI). Benefiting from array-CCD-based detection, the method enables spatially resolved imaging of defect distributions across a millimeter-scale field of view, offering higher efficiency and spatial resolution than conventional photothermal techniques. The consistency between thermally-induced deformation and photothermal absorption was validated using ion-implanted samples. An optimal LTP power density of 520 W/mm2 was identified. Four fused silica samples with different surface treatment levels were examined, revealing that improved defect control reduced deformation values, with a correlation coefficient of 0.99 at a fluence of 11 J/cm2. Finally, absorptive defect distribution models were constructed, showing that variations in defect type and distribution lead to distinct deformation behaviors. The resulting defect distribution models revealed distinct deformation behaviors associated with absorption centers, demonstrating the capability of LTP-DMI for sensitive and efficient assessment of ultra-low absorption features.

High-content imaging assays quantify cellular responses to chemical and genetic perturbations, yet continuous trajectories of individual cells are unobservable because cells are chemically fixed at acquisition. Perturbation modeling therefore reduces to inferring stochastic transport between control and treated populations observed only as separate marginals. While recent generative models achieve strong end-point alignment, boundary consistency does not determine intermediate evolution: multiple stochastic processes may connect identical marginals while traversing regions unsupported by observed single-cell morphologies. We introduce FreeBridge, a Schrödinger Bridge formulation for single-cell transition modeling under endpoint-only supervision. FreeBridge defines atomic states as instance-segmented single-cell representations, establishing a fixed cellular manifold, and learns stochastic transport constrained within this geometry via empirical latent support regularization. Across BBBC021, RxRx1, and JUMP, FreeBridge maintains competitive or improved endpoint fidelity and mechanism-of-action retention under a unified evaluation protocol; on BBBC021, it further reduces intermediate support violations. These findings highlight the importance of geometric grounding for biologically interpretable perturbation dynamics. Project page: https://y-research-sbu.github.io/FreeBridge/.

Structured light with expanded degrees of freedom has become an important driving force for advanced optical manipulation in biology, medicine, and chemistry. However, achieving synchronized control over particle multiplicity, precise spatial positioning, specific rotational states, and stability against rotational Brownian motion remains an unresolved challenge for structured optical tweezers. To address this challenge, we present a multifunctional optical trapping platform based on modulated trigonometric vortex beams (MTVBs). A petal-like optical trap is formed with tailored nonlinear phase gradient, which provides stable, predictable particle confinement forces. Precise regulation of optical torque, namely programmable rotational control, is achieved by modulating the parameters in MTVBs. The trapping mechanism integrates gradient-force-based positional locking with restorative torque generated by anisotropic intensity distributions, enabling highly stable confinement against both translational and rotational fluctuations. Dynamically sculpting MTVBs further permits advanced manipulations, including cluster aggregation/dispersion, controlled rotation, and precise spatial positioning. With its inherent flexibility and tunability, the MTVB platform establishes a versatile foundation for next-generation optical tweezers and sophisticated manipulation schemes.

A key challenge in digital holographic microscopy is the rapid and accurate estimation of the optimal in-focus distance without resorting to computationally intensive numerical reconstruction. Conventional autofocus approaches typically rely on a metric of sharpness by iterative scanning multiple positions along the axis of propagation, resulting in high computational cost and limited efficiency. In this work, we propose RegViT, a vision transformer-based regression framework that directly predicts the in-focus distance from a single hologram. By formulating autofocusing as a regression problem, the proposed method eliminates the need for iterative reconstruction. Experimental results on both synthetic datasets and real holograms demonstrate that RegViT provides an efficient single-shot estimate of the reconstruction distance and achieves competitive reconstruction quality under the tested off-axis DHM configuration. (c) 2026 Optica Publishing Group. All rights, including for text and data mining (TDM), Artificial Intelligence (AI) training, and similar technologies, are reserved.

Distributed acoustic sensing (DAS) based on phase-sensitive optical time-domain reflectometry ( Phi -OTDR) is a groundbreaking technique, which can continuously detect external vibrations or sound fields. In our previous work, a new low-complexity modulation method for the Phi -OTDR system was proposed to utilize the phase-shifting technique based on an unbalanced Michelson interferometer (MI-PST). However, it lowers the sampling rate to one-third of the pulse repetition frequency due to its working mechanism. In this work, we propose an intensity-cross-combination (ICC) method to efficiently recover the reduced sampling rate. In this method, the initial phase shift is recycled in a sequence of 0, 2 pi /3, and 4 pi /3 so that the phase signal can be calculated using the three-step phase-shifting method immediately after collecting one new phase-shifting light intensity. This new methodology addresses the limitation of having to collect all three different phase-shifting light intensities per signal point to calculate the phase signal in the original phase-shifting method. Experimental results demonstrate that the improved MI-PST-based Phi -OTDR system achieves the background noise level of 1.16 & times;10-3 rad/ root Hz under the sampling rate of 20 kHz and successfully recovers disturbance signals with frequencies ranging from 1 to 5000 Hz. This approach is both cost-effective and structurally simple, providing an alternative solution for the application of Phi -OTDR in perimeter security, urban transportation, and other fields.

Optical manipulation using structured light has become a cornerstone technique, propelling the development of laser-based applications in diverse fields such as biology, medicine, and chemistry. Conventional optical tweezers researches predominantly explore trapping multiple particles simultaneously. To meet the demand for asynchronous manipulation of multiple particles, a versatile optical trapping platform based on the open vortex beams (OVBs) is presented that utilizes their intensity gradients to achieve partitioned manipulation of both high- and low-refractive-index (HRI/LRI) particles. The particular optomechanical effects of proposed beam are demonstrated by three groups of the experiments including guided particles motion along programmable trajectories, Newton's cradle-like collective dynamics, and size-selective particle sorting. For HRI particles, sequential momentum transfer results from the combined action of an azimuthally modulated intensity gradient force and the particles' inherent inertia, which together constitute a previously unexplored coupled transmission mechanism. Meanwhile, the sorting of the LRI particles is completed by a differential response to optical forces. Building on this foundation, we further investigate the optical field properties and manipulation capabilities of OVB arrays, demonstrating their reconfigurability into arbitrary predefined modes for complex manipulations of particles. This proposed approach provides a versatile and efficient platform for targeted particle manipulation and separation, while also providing a new insight into optomechanical coupling in structured light fields.

Three-dimensional localization methods are crucial for accurately resolving particle positions and trajectories in complex systems, enabling deeper insights into dynamic processes in diverse scientific and engineering fields. Owing to its compact configuration and high space-bandwidth product, in-line digital holography has been widely used for three-dimensional particle localization and tracking. However, the accuracy of axial localization is often compromised by intrinsic twin image artifacts, leading to ambiguities in the reconstruction process. Additionally, multi-plane iterative reconstruction methods increase the complexity and computational efficiency of imaging. To address these two challenges, we propose a defocus-encoded coherent framework that simultaneously encodes defocus direction and distance. During holographic recording, a vortex phase with topological charge l=1 is introduced in the frequency domain, encoding the defocus direction into spiral interference patterns in the hologram. During numerical reconstruction, a digital vortex interferometry encoding is then used to quantitatively map the defocus distance to the rotation angle of the vortex pattern, enabling precise retrieval of the axial offset. Experiments and simulations demonstrate that the strategy enables volumetric imaging within a 10 mm3 field with precise axial resolution. The single-shot and non-iterative characteristics of this method enables robust and efficient tracking of dynamic particles.

Abstract Biomedical image segmentation is a fundamental problem in computational biomedicine that aims to precisely delineate anatomical and biological structures, tissue types, or pathological regions in biomedical images. Accurate segmentation is essential for interpretation, decision-making, and quantitative analysis across a wide range of biological and medical applications. Over the past decade, the field has undergone a profound paradigm shift, evolving from task-specific specialist models to universal foundation models. This review provides an in-depth analysis of the evolution, tracing how the limitations of local discriminative learning drove the transition toward transformer-based global modeling, and large-scale generative pre-training. To help navigate the diverse landscape of interaction paradigms, we introduce the first systematic taxonomy of promptable biomedical image segmentation, categorizing existing methods into six distinct types, enabling users to intuitively select appropriate prompting strategies based on visual demonstrations and quickly pinpoint relevant literature ( Prompt Type Visualization ). Beyond model architectures, we discuss parallel advancements in dataset development, evaluation protocols, and application-specific adaptations across radiology, pathology, and biology. Integrating these powerful foundation models with rigorous domain-specific adaptation has great potential to improve patient outcomes and healthcare efficiency. Finally, we highlight key challenges in trustworthiness and clinical integration that must be overcome to realize the potential of the next generation of biological and medical generalists.

High-precision aspheric surface testing follows the null-test principle, in which the computer-generated hologram (CGH) is an indispensable compensating element for aspheric null testing. In aspheric null interferometry based on a null CGH and a Fizeau interferometer, the measurement accuracy is significantly affected by fabrication errors of the CGH. Absolute testing achieves higher measurement accuracy than conventional aspheric null interferometry by separating the fabrication errors of the CGH. In this paper, an absolute testing method based on transmissive Twin-CGH is proposed for rotationally symmetric convex aspheric surfaces. By fixing the position of the Twin-CGH during absolute testing, the wavefront error introduced by the Twin-CGH and transmission sphere (TS) can be separated from the surface figure error of the aspheric mirror, thereby avoiding errors caused by CGH pose adjustment. Compared with the conventional transmission-type Twin-CGH method, this approach eliminates the need to precisely position a standard spherical reflector at the conjugate focus of the spherical wave. It bypasses a critical bottleneck: the conflict between conjugate positioning requirements in confined optical paths and CGH manufacturability constraints. Instead, it leverages high-precision lithography techniques such as electron beam lithography and laser direct writing to keep the pattern distortion of the Twin-CGH at a low level, offering high feasibility. To verify the validity and reliability of the proposed method, a Twin-CGH and a null CGH are fabricated separately. The proposed method was validated through two independent approaches: absolute testing using the Twin-CGH, and error separation via the N-position method with a null CGH. The results obtained by the two methods verify each other, which fully demonstrates the correctness and reliability of the proposed testing method. It is pointed out that the practical design of the Twin-CGH needs to comprehensively consider the spatial distribution of the 15 diffraction combinations listed in Table 3 at the aperture

High-frequency information fidelity in wavefront propagation directly determines the performance limits of computational optical systems such as holographic imaging and quantitative microscopy. However, the conventional angular spectrum method (ASM) faces two intertwined challenges in the discretization process: spectral aliasing and circular convolution errors. Existing improvements adopt fixed processing strategies that lack adaptability to diverse source spectra, making it difficult to balance computational accuracy and efficiency. Here, we propose an effective-information adaptive angular spectrum method (EIASM) based on the spectral statistics of the source field. By analyzing these statistics, EIASM dynamically constructs a frequency-space co-optimization mechanism that suppresses aliasing and eliminates wrap-around errors while achieving computational complexity proportional to the effective information content of the source field. Simulation results demonstrate that at a propagation distance of 1000 mm, the computation time of EIASM is reduced to 1% of that of the wide-window angular spectrum method (WWASM), while maintaining comparable accuracy to WWASM under large numerical aperture conditions. The proposed method can be extended to the matrix-product angular spectrum framework, alleviating its irreversibility and computational redundancy issues, and provides a high-precision, high-efficiency adaptive propagation paradigm for forward and inverse modeling in computational imaging.

Optical Diffractive Tomography (ODT) is widely used to reconstruct the 3D refractive index distribution of weakly absorbing and scattering samples. This is achieved by solving the inverse scattering problem, which mathematically relates the measured scattering field to the object's properties. However, the recovery of the scattering field information relies on multiple recordings. To satisfy the requirements of practical applications, such as living cell imaging and real-time material characterization, real-time ODT is desirable. Here, we propose a single-shot ODT technique based on polarization multiplexing. The scattering information under different illumination angles are encoded using multiple polarization states. The polarization camera and intensity extraction matrix are used to decode the information, achieving simultaneous recording and decoding of multiple information without overlapping. The Kramers-Kronig relations is used to reconstruct the scattered field information of the sample and further applied to the 3D refractive index reconstruction of the sample. The multi-angle recovery of the scattered field can not only realize the 3D refractive index reconstruction, but also ensure the improvement of the lateral resolution. In addition, iterative processing, axial scanning, and deconvolution can be avoided by using Kramers-Kronig relations. Experiments are conducted to demonstrate the effectiveness of this method. Phase-type USAF1951 test target and un-label onion cells are used in experiment, the experimental results demonstrated that the proposed method represents a significant advancement both in resolution improvement and real-time performance in the 3D RI measurement of samples, which achieve lateral resolution of 770 nm and an axial resolution of 8 mu m by using a microscope objective with a numerical aperture (NA) of 0.4. We expect the effective application of this technology in real-time detection fields such as live cell imaging and surface processing, so as to accelerate the dynamic observation and analysis of related processes.

Twin-CGHs enable high-precision absolute aspheric testing, yet they suffer from severe stray wave interference. In this work, we fabricate three phase-type Twin-CGH prototypes using standard GDSII layout files and analyze the effects of different phase binarization strategies on their stray wave suppression performance. The results reveal that, in the binary phase design of Twin-CGH, any deviation or fabrication-induced drift of the phase duty cycle away from 0.5 will introduce amplitude modulation into the diffracted wavefield. This produces a significant energy rise for zero-order diffraction, and the wavefronts governed by the extra phase term exp [ i ( β±α )], while simultaneously reducing the diffraction efficiencies of the desired spherical and aspherical reference wavefronts. Drawing on these conclusions, we propose a universal design strategy for Twin-CGH. This method can avoid the adverse impact of stray light generated by extra phase functions on high-precision testing, and can ensure the energy stability of the reconstructed aspherical and spherical waves.

Free-space optical (FSO) communication holds significant potential for future systems due to its high bandwidth and installation-free advantages. However, FSO faces critical challenges including atmospheric turbulence and security vulnerabilities such as susceptibility to interception. To address these issues, this paper proposes a multi-parameter encrypted FSO communication scheme based on generalized polygonal non-diffractive vortex beams (GPNBs). By incorporating multiple structural dimensions with edge count factor, smoothness factor, and topological charge. The system achieves multidimensional information modulation and enhanced encryption capabilities. While a MobileNetV3-Smallbased neural network decoder enables high-precision demodulation. Experimental results demonstrate successful transmission of three color images over 1 m with a bit error rate (BER) of 3.9 & times; 10(-3) using combinations of 8 orbital angular momentum (OAM) modes, 3 smoothness factors, and 3 edge number factors. The non-diffractive characteristics maintain BER at 5 & times; 10(-2) under moderate to high atmospheric turbulence, while the beam's self-healing properties enable recovery from obstructions within 1 m. This multidimensional encryption mechanism and lightweight neural network decoding strategy provide a promising solution for high-capacity, secure, and robust FSO communication in complex environments.

Differential absorption lidar (DIAL) is widely used to monitor spatial and temporal variations in the distribution of atmospheric gases, such as CO2 and O3. Single-photon detectors are critical components of DIAL systems, and their after pulse effect can distort echo signals, significantly compromising the accuracy of gas concentration inversion. This study introduces an intelligent optimized after pulse correction algorithm that utilizes differences in signal response between two detectors. The algorithm evaluates the after pulse characteristics of the detectors and applies real-time corrections during signal processing. This effectively mitigates the interference caused by strong after pulse effects on signal integrity. Comparative experiments with real lidar signals demonstrate that this algorithm substantially enhances the detection performance of DIAL systems. Notably, the proposed method enables online evaluation and calibration of the after pulse effect using lidar signals. Unlike traditional calibration approaches that rely on pre-installation detector testing, this technique provides a more thorough evaluation and effectively addresses the worsening of after pulse effects due to aging lidar system components. The results show that this method increases the maximum nighttime detection range of the differential absorption lidar from 1500 m to 2000 m.

Weak light intensity positions induced by interference fading adversely affect the sensing performance of phase-sensitive optical time-domain reflectometry (Φ-OTDR). Most effective fading suppression methods rely on frequency or phase modulation of the light source, which requires complex hardware modifications. To solve the above issue, this paper proposes a novel multi-channel data synthesizing method based on deep neural network (MDS-DNN) to reduce the impact of interference fading on the signal-to-noise ratio (SNR) of Φ-OTDR. The proposed algorithm can work efficiently without any modification of the conventional Φ-OTDR setup. The spatial sampling rate of the Φ-OTDR systems is typically much higher than the spatial resolution. This means that neighboring sampling points carry the same external vibration signal, providing redundant information. Therefore, it is possible to perform comprehensive analysis on these multi-channel data to improve the suppression capability of interference fading noise. This work designs a long short-term memory (LSTM) network-based framework and an end-to-end training strategy to automatically learn the correlation between these multi-channel data and the ideal sensing signal. Simulation and experimental results show that the MDS-DNN algorithm can effectively suppress phase noise and improve the SNR at fading positions. Experiments using the data collected from the actual Φ-OTDR system demonstrate that the output SNR can reach 49.88 dB, which is 19.65 dB higher than the average level of the input channels. Moreover, the MDS-DNN method reduces the false alarm rate caused by interference fading by one order.