Nonlinear optical microscopy is a vital technology in biomedical imaging and neuroscience. As multi-modal imaging significantly enhances the diagnostic utility, its conventional implementations rely on complex multi-laser configurations that limit the accessibility and upgradability of existing systems. In this study, we develop a compact broadband ultrafast all-fiber laser source enabling simultaneous four-modal nonlinear imaging, including two-photon fluorescence (2PF), second-harmonic generation (SHG), three-photon fluorescence (3PF), and third-harmonic generation (THG). To showcase its potential, we conducted high-quality multi-modal imaging on various biological samples, including mouse brain sections, mouse kidney sections, melanoma, oral tumors, and breast tumor tissues. This robust all-fiber laser source offers a simplified yet powerful solution for label-free structural and molecular analysis in complex biological systems.

The annual thyroid cancer incidence has been increasing. Thyroid cancer is categorized as a malignant neoplasm within the endocrine system. Fine-needle aspiration cytology remains the benchmark for thyroid cancer detection; however, the accuracy of the procedure depends on the practitioner's expertise. Numerous challenges are associated with this process, such as obtaining inadequate cellular samples, mispuncturing the target lesion, and collecting nonrepresentative cell samples. These issues hinder proper cellular evaluation and increase the likelihood of misdiagnosis, ultimately impacting patient outcomes and the treatment trajectory. This research primarily aims to improve the diagnostic accuracy of thyroid cancer by introducing an innovative computer-aided diagnostic tool that leverages advanced deep learning techniques. Second-harmonic microscopy enables the fine extraction of morphological characteristics of collagen fibers within thyroid tissues, revealing significant differences in the distribution and organization of collagen fibers between normal and malignant tissues. In this study, we quantified the morphological alterations of collagen fibers by initially analyzing second-harmonic generation (SHG) images through a collagen scoring system based on feature extraction via least absolute shrinkage and selection operator regression. Model efficacy was assessed using receiver operating characteristic curves. Furthermore, we classified normal and malignant thyroid tissues in the validation cohort through three distinct deep-learning architectures (Mobile Neural Networks Version 3 (MobileNetV3), Visual Geometry Group 16(VGG16), and Pyramid Vision Transformer v2 (PVTv2) in combination with SHG image data. Overall, MobileNetV3 achieved the best classification performance (87.4%). This study provides preliminary evidence for the effectiveness of deep-learning algorithms in differentiating between malignant and normal thyroid tissues. This significant advancement offers valuable technological support for detecting thyroid cancer in clinical environments and is expected to enhance both the accuracy and efficiency of diagnostic practices.

We explore laser-induced dewetting of thin gold films deposited on different underlayers consisting of optical cavities of gold and silica. Our observations indicate that the morphology and optical response of laser-exposed regions are sensitive to the optical environment in vicinity of the dewetting layer.

Photoconductive (PC) terahertz emitters operating at communication wavelength (∼1550 nm) offer particular advantages through compatibility with cost-effective and reliable fiber lasers. However, their performance is often hindered by high dark currents and inefficient photocarrier collection. This work introduces a bias-free PC terahertz emitter featuring an ITO/U-InAs/P-InAs tri-layer structure with a dielectric metasurface to overcome these limitations. The configuration achieves a more uniform optical field distribution in the U-InAs/P-InAs layers and a high optical absorption of 95% at 1550 nm by using the dielectric metasurface with electric and magnetic dipole mode degeneracy, which is performed using the Finite-Difference Time-Domain (FDTD) method. To effectively accelerate the photocarriers, a large built-in electric field is established across the tri-layer by leveraging energy band engineering and aligned with the optical field distribution. Our numerical calculations show that this synergy enables efficient photocarrier collection without external bias and obtains a high optical-to-terahertz conversion efficiency of 3.2% and a bandwidth of 3 THz. These advancements highlight the potential of our design for high-efficiency terahertz sources.

The development of noninvasive methods for diagnosing skin diseases is a major problem in modern medicine. This paper presents an advanced multimodal approach that integrates Optical Coherence Tomography, Diffuse Reflectance Spectroscopy, and High Frequency Ultrasound data. Image processing, spectral and texture analysis, and machine learning techniques, including gradient boosting and fusion algorithms for unimodal classifiers, were used to analyze the data. The developed multimodal model provides high accuracy (96%) in differentiating benign and malignant neoplasms, demonstrating significant potential for clinical diagnostics. These results highlight the importance of integrating data from multiple modalities to improve diagnostic accuracy and reliability for skin neoplasms.

Single-photon avalanche diode (SPAD) pixel scaling is essential to meet the increasing demands for high-resolution, compact, and power-efficient time-of-flight (ToF) sensing. In particular, the 3D-stacked approach enables aggressive pixel scaling by separating the SPAD and readout circuits into different wafers, thus maximizing the fill factor while minimizing the pixel pitch. However, pixel miniaturization often leads to degraded SPAD performance due to the premature edge breakdown (PEB) and the reduced number of photon-generated carriers that go through the avalanche multiplication region. In this work, we overcome these challenges by optimizing the doping profile to enhance the carrier collection in the device. We present a detailed analysis of the optimization progress by evaluating breakdown voltage (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">V_B</i>), dark count rate (DCR), and photon detection probability (PDP), highlighting the trade-offs and recovery achieved through successive doping refinements. The optimized device achieves a PDP of 37% and a timing jitter of 85 ps at 940 nm. Compared to prior 3D-stacked back-illuminated (BI) SPADs, our work exhibits one of the smallest pixel pitches to date, yet retains competitive PDP and jitter characteristics. This combination of aggressive scaling and robust performance positions the proposed SPAD as a promising solution for LiDAR, 3D imaging, and future wearable sensing systems.

Micropillar cavity lasers demonstrate operation up to 255 K. At this temperature, the lasing threshold and Q-factor, extracted at lasing threshold for a 4 μm diameter pillar are about 2 mW and 16800. For pillar lasers, the concept of which is based on an optical aperture, continuous-wave lasing is realized at room temperature (300 K). For a 15 μm diameter pillar with an aperture of 4 μm, single-mode lasing at 960 nm is demonstrated with a threshold power of about 12 mW and a Q-factor at threshold of 13000.

We present a comprehensive comparative study of passive quenching active recharge (PQAR) and active quenching active recharge (AQAR) analog front-ends (AFEs) designed for 64×64 pixel arrays of single-photon avalanche diode (SPAD) pixels. These AFEs were fabricated using a back-side illuminated (BSI) 3D-stacked 40 nm/40 nm CMOS image sensor process. While configured as an array, initial verification was conducted at the pixel level, comparing both AFE types. To enable a truly fair and direct performance evaluation, which was often difficult in prior independent studies due to the use of different devices, processes, architectures, or assumptions, both AFE types were fabricated under identical process, wafer, and layout conditions on the same die. The SPADs exhibit a breakdown voltage of 22.5 V, a low dark count rate (5.1 cps/μm<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> at 2 V excess voltage), and a high photon detection probability (25.1% at 850 nm at 2 V excess voltage). The AQAR pixel occupies a larger area (88.56 μm²) compared to the PQAR pixel (55.75 μm²), reflecting an area-performance trade-off. We analyzed methodologies for defining SPAD dead time, adopting inter-arrival time histogram analysis under intensive light. Experimental results demonstrate precise dead time control for both AFEs, achieving a minimum dead time of 2 ns and confirming maximum count rates up to near 500 Mcps. Moreover, this study experimentally verifies the effectiveness of active quenching in afterpulse suppression, leading to reduced noise and consequently contributing to an improved signal-to-noise ratio. The results of this study provide useful information on the design trade-offs and performance of PQAR and AQAR AFEs for various high-performance single-photon detection applications.

This study developed and validated a non-contact clinical system for myocardial ischemia detection using a fiber optic sensor (FOS) placed under a mattress. A novel algorithm was introduced to assess the hemodynamic consequences of ischemia by quantifying changes in cardiac function. The algorithm calculated the ratio between the IK and JJ intervals of the ballistocardiography (BCG) waveform, a temporal feature that serves as a biomarker for myocardial contractility impaired by ischemia. An initial cohort of 5 healthy subjects and 5 patients with myocardial ischemia was used to establish a diagnostic threshold, yielding an area under the receiver operating characteristic (ROC) curve of 0.9556. A separate clinical validation with 5 additional subjects subsequently confirmed the system's effectiveness. The analysis demonstrated that an 8-second signal duration was sufficient for accurate detection, and processing times ranging from 1.2 ms to 30.7 ms across various hardware platforms affirmed the system's real-time capability. The results highlight a robust and non-invasive system for continuous cardiac monitoring, offering an automated and objective tool to support the early clinical detection of myocardial ischemia.

Millimeter-wave systems require absorbers with high absorption efficiency, broad bandwidth, and multi-functional tunability. However, existing devices based on composite structures and materials face design challenges: traditional methods hardly achieve all three indicators simultaneously, often sacrificing one performance for another, making "high absorption, broad bandwidth, multi-function" integration highly difficult. To overcome this fundamental trade-off, this study designs a W-band (75-110 GHz) composite absorber, integrating VO2 and graphene, and combining periodic resonant structures with a backplane-free phase-change wire-grid architecture. When integrated into the periodic resonant structure, the VO2 wire-grid provides thermal ‘on-off’ switching of absorption, whereas graphene offers electrical tunability— both achieved by modulating the effective surface conductivity and thus the impedance matching condition. Guided by an XGBoost-based closed-loop inverse design framework, the XGBoost model optimizes parameters, with convergence achieved after four closed-loop simulation iterations. Results show the VO2 wire-grid absorber performs excellently at 90° incidence: it achieves near-perfect absorption (> 99%) over a 21 GHz sub-band (82-103 GHz), showing an 80% improvement in average absorption compared to copper(Cu) backplanes, under identical unit-cell geometry and simulation conditions, with each configuration independently optimized for peak performance. Graphene provides a maximum modulation amplitude of depth of 68%. Innovations of this study include constructing a periodic resonant wire-grid structure that abandons traditional backplanes and uses wire-grids to expand loss areas, enabling dual-mode multi-physics control through co-integration of VO2 and graphene, and establishing a machine learning closed-loop framework for parameter optimization. It explores a new paradigm for multi-functional tunable absorbers, providing high-performance solutions for MMW systems such as radar, communications, and biomedical imaging.