Session Index

To find out how neurons in Suprachiamatic nucleus, a small nucleus at the bottom of the hypothalamus that acts as the principal circadian clock of brains, generate a coherence daily rhythm, the computed neuronal network at SCN is necessary to be clarified. Among genetically encoded voltage indicators, ASAP-family are currently the only ones to reflect single-spike signals and to optically characterize individual neuron coding dynamics. We report a 1.5-kilohertz-frame-rate microendoscopy system with GRIN lens which enables high-speed two-photon voltage imaging to observe the quick response of action potential in deep mice brain, and thus help identify the millisecond-level neuronal activity.

  Preview abstract

Harmonic generation microscopy (HGM) is a non-invasive in-vivo label-free 3D imaging technique. In recent years, HGM shows great potential in improving quality and efficiency in medical treatment. However, the image modality of HGM is different from widely-used H&E-staining approach which pathologists are familiar with. This paper proposes a new image translation from HGM images to H&E-stained images. Specifically, this paper presents an unsupervised deep-learning based methodology to effectively synthesize H&E-stained image with given HGM image. Result indicates that the proposed methodology is promising in medical image translation and hopefully will facilitate adopting HGM in clinical workflows in the near future.

  Preview abstract

The temporal focusing multiphoton microscopy can rapidly provide 3D imaging, but strong scattering through biotissue will degrade the image quality and reduce the penetration depth. In the study, we present a prediction learning model with depth information to overcome. A 3D U-Net network with digital propagation matrix has been developed to predict and improve different layer images under cross-modality training. Furthermore, a long short-term memory-based network, which is designed to forecast the deeper information according to previous 3D information, is introduced for the prediction of depth information.

  Preview abstract

This study explores and implements the use of U-curve method to determine the regularization parameter to reconstruct diffuse optical images and use some simulated measurement data to observe the image reconstruction. Some synthesized case examples and experimental ones as well are conducted to verify the performance. uses some simulated measurement data to observe the image reconstruction.

  Preview abstract

Isotropic Quantitative Differential Phase Contrast Microscopy (iDPC) is a novel
phase imaging method. We can accurately quantify structural information in living cells under
label-free conditions. Here, we present a U-net model that implements the transformation from
intensity to phase images and obtain the quantitative phase information. Compared to use of
Tikhonov regularization, this deep learning model not only replaces complex calculation and
parameter setting but also achieves isotropic results with only half the number of images, and
incresese the imaging efficiency two times. The results facilitate time-lapse observation for
recording dynamic cellular information.

  Preview abstract

We present a label-free approach to achieve three-dimensional measurement and quantitative analysis in tens of seconds for cell death using holographic tomography. In the experimental results, the morphological changes and internal structures of neuroblastoma cell in autophagy and apoptosis are demonstrated with three-dimensional tomographic images.

  Preview abstract

This study aimed to develop a versatile OCT ophthalmic detection platform, including structural imaging, angiography, and stimulus-evoked Intrinsic Optical Signal (IOS). We obtain cross-sectional images of the retinal structures of BALB/c mice under optical stimulation by OCT scanning. In addition, IOS assessments were performed for the cross-sectional images of retinal structures, providing a non-invasive, non-contact, and relatively simple method for evaluating eye function.

  Preview abstract

PS-OCT is a non-destructive and three-dimensional imaging technique that can provide polarization property, e.g., phase retardation and optical axis, as well as the architectural information similar to conventional OCT from the sample. In this study, we have developed a high-speed PS-OCT imaging engine by using a novel wavelength-swept laser light source based on HCG-VCSEL. Example PS-OCT imaging including the human fingernail junction, 3D plastic printing material and the chicken breast tissue demonstrated the depth resolved measurement of the multifunctional information of the sample with PS-OCT and HCG-VCSEL light source at an A-scan rate of 250 kHz.

  Preview abstract