Session Index

In this study, we have developed a catheter-based OCT system allowing volumetric imaging of the cervix. The rotary junction and catheter are custom-designed to tailor the aforementioned biomedical applications. Preliminary imaging results of the human fingertip were shown to demonstrate the imaging capability of the developed imaging system.

 

We have developed a swept-source OCT system using two 1060 nm wavelength-swept lasers with a sweep rate of 100,000 and 200,000 A-scans/sec to reconstruct optical coherence tomography angiography (OCTA) images. Preliminary results of the OCTA images of the mouse ear skin are presented in this paper.

 

We show that the epithelial-mesenchymal transition (EMT) factor, Snail, promotes the invasion capability of head and neck squamous cell carcinoma (HNSCC) cells by increasing cell traction forces and decreasing the nuclear stiffness, associated with the increment of cell focal adhesion and decrement of lamin B1 expression.

 

Light absorption is common in biological specimens which potentially serves as a unique contrast mechanism for microscope imaging especially when detecting non-fluorescent targets. We present a wide-field photothermal imaging technique that measures the local heat through interference and demonstrate absorption imaging of single metallic nanoparticles, as small as 5 nm.

 

We develop a deep learning algorithm as an alternative to the existing Tikhonov regularization, relying on the idea of sensor-domain to image-domain transformation. U-net with skip connections is used to extract the features and obtain the contrast image. The output is obtained by multiplying contrast images with the background coefficients.

 

The performance of interferometer for optical coherence tomography (OCT) such as flatness and broadband are important to maintain the axial resolution and sensitivity. We develop a broadband coupler using numerical optimization to obtain a broadband flat response with more than 180 nm bandwidth from a transmission and multiple reflection.

 

We have developed a multi-scale long-wavelength optical coherence microscopy imaging system allowing three-dimensional imaging of biological tissue. Besides, we demonstrated the ability of our self-developed automatic mosaic stitching algorithm.

 

Precise retinal layer segmentation in optical coherence tomography (OCT) helps early clinical diagnosis of some diseases. In this work, we established a segmentation algorithm for the retinal layer boundaries based on deep learning. Our method successfully segments three retinal layers with acceptable errors compared to traditional methods.

 

Here, we demonstrated a label-free, sub-micron multi-harmonic generation microscopy to elucidate the histopathological hallmarks of Alzheimer’s disease human brain ex vivo. Our harmonic generation microscopy has shown the capability to differentiate simultaneously structures of brain viz. axons, soma along with AD pathological hallmarks: Aβ-plaques and neurofibrillary tangles.

 

Diabetic neuropathy is diagnosed by skin biopsy to quantify intra-epidermal nerve fiber. Our recent study indicated the capability of third-harmonic-generation microscopy (THGM) to image intra-epidermal free-nerve endings. In this study, we successfully apply the stain-free trauma-free THGM to in vivo evaluate epidermal innervation by quantifying the number of nerve fibers.

 

Optical sectioning in our system is achieved by HiLo imaging technique. High speed hybrid illumination for background elimination is generated by a digital micro-mirror device (DMD),with electrically tunable lens (ETL) as a mechanical-scan-free device. Our system offers multiple advantages, which may enable large scale dynamic imaging of biological tissues.

 

We demonstrate the first observation of the intra‐epidermal free-nerve-endings (IFNEs) with histopathology details by third-harmonic-generation on ex vivo human skin. With the high image contrast, noninvasiveness, high resolution, high penetration of harmonic generation microscopy, we aim to label-free slide-free in vivo image IFNEs for the peripheral neuropathy diagnosis.

 

The advantages of the lattice light-sheet microscopy in controlling the light field are applied in optogenetics. It can be obtained that the Bessel beam reflects different optogenetic response areas under different parameters. The Final we can produce two independent membrane ruffling reactions in the same cell at the subcellular scale.

 

The mask edge detection algorithms using Sobel, Prewitt, and Canny operation are performed for the one-dimensional scanning projected fringe profilometry. The optimized mask detection algorithm helps to identify the amplitude of the projected fringes, to build the amplitude of the in-focus image and the 3-D profile of the inspected object.

 

Pharmaceutical approaches to treat dementia related sleep disorders have failed. Light therapy alone shows promise. A 24-hour dynamic lighting system with IoT control and validation has been developed and is undergoing final reliability testing. Upon final IRB approval, patient recruitment will begin.

 

The surface plasma resonance (SPR) phase sensitivity is proposed to utilize two Mach–Zehnder interferometers in parallel, a laser source to detect SPR effect and broadband source as a reference point, to gauge interferogram displacements. The displacement variation of different miR-21 concentrations demonstrates the sensitivity as 0.47 rad/μM.

 

Neonatal jaundice is a common condition in Asian neonates, and the existing transcutaneous bilirubin meters are not reliable enough because of inaccurate empirical models. We have developed a handheld diffuse reflectance spectroscopy system with an advanced model. The performance of our handheld system was successfully validated with a phantom study.

 

We have developed a polarimetric imaging system to obtain full Mueller matrix images within ten-odd seconds. In this work, a specially designed diffractive multifocus grating was installed in the Mueller matrix imaging system to extend the system with three-dimensional imaging ability of the full polarization characteristics.

 

Acquiring high-dimensional (HD) optical information in real-time is a challenging task. Spatial-spectral filtering properties of the volume holographic grating offers unique advantages for obtaining simultaneous multidimensional information without transverse or axial scanning. Here, we discuss the state-of-the-art volume holographic microscope and endoscope system for label-free as well as fluorescence imaging.

 

We developed (20-21) semipolar micro light-emitting diodes (-LEDs) as RGB pixel display with quantum dots photoresist (QDPR) and Visible Light Communication (VLC) application.

 

Electric field Monte Carlo method can be used to include the phase and polarization state of complex light in turbid media. In this paper, based on EMC framework we investigate the depolarization dependency of special beams over the penetration depth and the scattering properties in turbid medium for biomedical applications.

 

Blood-related diseases can cause red blood cell abnormal morphology and deformability. The high-throughput label-free digital holographic microscopy (DHM) with the refocusing feature, can be used to record cells’ three-dimensional refractive index (RI) distribution and viscoelastic. Via simulation, the relation between Zernike fitting and phase image has been discovered and implemented.

 

The mammalian cochlear is a complex anatomy related to the hearing function and especially its spiral structure and hair cell in the inner ear. In this study, we quantified the mouse cochlear’s optical properties through the spectroscopic comparison, two different center wavelength home-build swept-source optical coherence tomography were using.

 

Laser material processing at micrometer resolution includes optical ablation with high laser pulse energy and laser-induced material modification. According to the material optical properties, the multiphoton-excited fluorescence microscopy (MPEFM) could perform noninvasive inspection by reconstructing the 3D fluorescence images that could directly see the processed morphology in specimen

 

Graphene was shown to be a useful additive material to SERS substrate in DNA detection thanks to its ability to significantly enhance the DNA signal, while the sample preparation is very fast and simple. Therefore, the addition of graphene could offer a new DNA detection method using nano-sized cells.

 

In this study, we proposed to use optical coherence tomography (OCT) to investigate the skin tumor. Then, U-net model is implemented for automatic identification of tumor region. From our results, the proposed method enables to effectively distinguish different tissue components of skin and the tumor region can be accurately distinguished.

 

In this study, a novel compact liquid assay detection platform based on gradient grating period guided-mode resonance (GGP-GMR) atop of a CCD is demonstrated. The system successfully detect the intensity and peak wavelength variation at different concentration for two commonly used fluorescent dyes (6-FAM and 5-TAMRA) in DNA detections.

 

We demonstrate a portable and multichannel SPR biosensor based on hand-held platforms. We design a disposable chip that includes microfluidic and micro-optical system. The overall size of the portable system is only 15cm× 20cm×10cm. This cost-effective and sensitivity SPR biosensor system for medical diagnostics, environmental monitoring, and food safety applications.

 

In this study, the non-invasive evaluation of reflective heart rate monitoring (HRM) is performed. The optical simulation is used to evaluate the capability of a reflective HRM with human skin. Thus, the choice LED and photodiode (PD)、baffle plate, the distances between the components are important for HRM design.

 

The GaN structures with and without InGaN QWs were fabricated by the MOCVD to serve as a SERS substrate. The Raman enhancement of the single-strand Cytosine DNA on theses substrates showed the effect of InGaN QWs in SERS, which can be understood by the 1D-DDCC software built-energy band diagram.

 

We’ve reached a significant breakthrough in Fabry-Pérot spectrometer, as the spectrum must be measured less than the limit of a free spectral range and constructed a new form of spectrometer: Experimentally, we achieved the scanning performance of the FP spectrometer by the change of the refractive index through driving voltage.

 

In this study, a novel gradient grating period guided-mode resonance (GGP-GMR) biosensor is proposed, and albumin was successfully detected as a test model. The simple design and easy fabrication of the GGP-GMR sensor combined with its simple readout and optical path design facilitates system miniaturization.

 

To reveal the changes of spruce due to chemical treatment and aging, a two-photon hyperspectral system has been applied to investigate the autofluorescence and second harmonic generation within wood. The cellular structures were observed through optical sectioning and the spectral variations were revealed among different samples and different cellular structures.

 

We demonstrated a miniaturized sensor to detect the urinary urea concentration in cats. The sensor composed by pH indicator doped hydrogel and a Y-shape fiber that delivered and collected signals. A linear calibration curve was successfully built to read the urea concentration in urine solution obtained from cat litter.

 

An exploratory dermatology clinical trial was conducted using cellular-resolution optical coherence tomography (OCT). The nuclei in skin OCT images were segmented by the convolutional neural network. After segmentation, the achieved Dice coefficient was 0.58±0.02. The depth-dependent nuclei ellipticity can be readily obtained from the in vivo measurement.

 

Quantitative differential phase contrast (qDPC) is a label-free imaging technique for resolving the phase information of the thin transparent object. To improve qDPC system for monitoring rapid changes in the sample. We adapt deep neural networks (DNN) to achieve isotropic differential phase contrast (iDPC) imaging and reduce the acquisition time.

 

Sleep disturbances accompanying Alzheimer’s disease are a serious problem. A thermal imaging camera, combined with a passive infrared motion sensor controlled by an ESP32 were used to monitor a bed, in order to provide feedback on patients' motion and bed occupancy.

 

This study aims to compare the optical clearing effects of polyethylene glycol 400, propylene glycol (PG), and glycerol with the combination of ultrasound treatment through human skin in vivo. Our results indicate that 25% PG with ultrasound treatment for total one hour provides optimum efficiency among all conditions.

 

The 3D dendritic path of corneal sub-basal nerve makes it difficult to visualize. Using the strong scattering of the Bowman’s layer as a guide, the sub-basal nerve plexus is projected on a 2D plane to extract the features. As a result, 12 parameters are defined to quantify the nerve morphology.

 

OCT was integrated with Raman spectroscopy to distinguish various cancerous and normal skin cells via machine learning. Features of OCT images were effective in differentiating the cancerous and normal cells (~80% accuracy). Using the Raman spectra, melanoma cells and keratinocyt-based cell carcinoma can be distinguished with near 100% accuracy.

 

Migraine is a common neurological disease and based on the cerebrovascular reactivity in prefrontal cortex, we used near-infrared spectroscopy (NIRS) as a biomarker during mental arithmetic. In our study, we found out significant differences between healthy controls and migraine patients, and the classification results by machine learning had 85% accuracy.

 

Two kinds of data: white-light (WLI), narrow-band (NBI) spectrum data from their endoscopic images, can be divided into four categories, Normal, Dysplasia, Dysplasia-ECA and ECA. Then, trained through the Single Shot Multibox Detector (SSD) with convolutional neural network (CNN). The precisions corresponding to stages are 85%, 84%, 88%, and 92%.

 

Saturated excitation microscopy (SAX) can provide high spatial resolution images with the help of higher order harmonic frequencies. However, obtaining images at different depth is a challenging task. Here, we show multi-depth imaging in SAX microscope using tunable focusing lens without moving samples or microscope objective.

 

In this study, we have developed a high-resolution two-photon fluorescence microendoscope, based on GRIN lens, to observe longitudinally the neuronal activity in the deep brain of awake mice. Furthermore, with ultrahigh 2 kHz frame acquisitions, the neuronal activity of diseased brain vs. normal brain can be easily differentiated.

 

Temporal focusing multiphoton excitation microscopy is potential to provide fast bio-imaging. Its surpassing acquisition rate is better than extensively used two-photon excitation microscopy, but the scattering phenomenon induces much high noise. To improve image quality, we induce deep learning method and present cascade 3D U-Net to realize high-speed 3D imaging.

 

Pleural fluid cytology serves as useful information in diagnostic lung cancer. In this study, we propose a model for analyzing pleural effusion cytology images based on hyperspectral imaging technology. The different hyperspectral spectra features are studied. We expect the proposed method could assist doctors in detecting lung cancer rapidly.

 

In this study, we use a 3D U-Net to improve the axial resolution of rapid volumetric multiphoton microscopy without reducing volume rate. Currently, 30 volume rates can be kept with high image quality.

 

An overview of recent progress of the author’s multi-contrast optical coherence microscopy (OCM) is presented. This OCM is based on holographic signal processing, time signal analysis, and Jones matrix imaging. The technical details of each technology and applications for in vitro tissue imaging are shown.

 

I introduce the Intelligent Image-Activated Cell Sorter 2.0, a technology that performs deep-learning-assisted image-based sorting of single live cells from large heterogeneous cell populations on a microfluidic chip with a high throughput of >2000 cells/s, enabling a new class of biomedical applications.

 

A pre-train image conversion scheme is adapted into an image classifier network of skin cancers. The network is separately trained on both OCT images before and after conversion into H&E-liked images. We compared the training results to anaylize factors which influenced the accuracy and discovered some insight of deep learning.

 

Utilizing pseudospectral time-domain (PSTD) simulation technique, we model the phenomenon of optical phase conjugation (OPC). Specifically, we investigate the feasibility to focus light with imprecise OPC information. Research findings may provide essential information to determine the contribution of certain factors in focusing light in turbid media.

 

The purpose of this research was aimed to evaluate the performance of a novel convolutional neural network (CNN) by automatically differentiation between benign and malignant entities of lung tissues during surgery with artificial intelligence-based optical coherence tomography (OCT). According to the results, our result shows excellent visualization of OCT-captured images.

 

Skin lymphoma is a rare skin disease. Easily misjudged as the more common atopic dermatitis or psoriasis. We use machine learning image recognition technology. I hope to use deep learning to achieve preliminary diagnostic capabilities, shorten the time between consultations, and quickly help patients treat illnesses.

 

Our photo targeting method can reduce the light scattering.We hoped that this targeting device will increase targeting accuracy, decrease operative time, and decrease the radiation from fluoroscopy.

 

The lossy mode resonance sensor with a self-heating electrode to immobilize bovine serum albumin (BSA) was proposed and studied. The sensor is a planar waveguide coated with a thin film of ITO. With controlling the external voltage to generate heat through ITO, the LMR wavelength shift is more obvious.

 

Extracorporeal membrane oxygenation (ECMO) is a lifesaving rescue treatment in refractory respiratory or cardiac failure. However, no guidelines or reliable clinical parameters are available for adjusting ECMO treatment in terms of machine settings. Therefore, this study applied near-infrared spectroscopy (NIRS) to observe these patients' global circulatory status.

 

Implantable retinal prostheses are now regarded as solutions to those who lose their vision due to retinitis Pigmentosa. To supply power to these prostheses, infrared light is illuminated to the photovoltaic cell inside them. Light Tubes are applied to guide the light. Results and discussion are presented.

 

The large-area characterization, precision, intelligent automation, and high-efficiency detection of nanostructures for 2D materials have not yet reached an industrial level. We propose deep neural network, one-dimensional (1D) convolutional neural network to explore the correlation between the accuracy of model recognition and the optical characteristics of few-layers MoS2.

 

Healthy work lighting is primary for indoor workers nowadays. An office lighting design system conforming to users’ comfort is therefore proposed in this study. For conforming the comfort, electroencephalogram (EEG) and heart rate variability (HRV) are regarded as the physiological parameters for objective evaluation, and questionnaire is the subjective evaluation.

 

The research developed a SWIRCNN architectures to reconstruct the human blood vessel of near-infrared image. The SWIRCNN was based on a de-noising convolutional neural network (DnCNN), which was used residual neural network (ResNet), and the SWIRCNN
network can remove the periodic noise from 19.1 dB to 26.4 dB.