|
16:00 - 16:15
|
Manuscript ID. 0965
Paper No. 2023-SAT-S0605-O001
|
| Syuan-Ruei Chang
|
Deep Temporal Focusing Multiphoton Microscopy Bioimaging Using Deep PhyCell and ConvLSTM
Hao-Chung Chi, Syuan-Ruei Chang, National Yang Ming Chiao Tung University (Taiwan); Yvonne-Yuling Hu, National Cheng Kung University (Taiwan); Feng-Chun Hsu, Chun-Yu Lin, National Yang Ming Chiao Tung University (Taiwan); Shean-Jen Chen, National Yang Ming Chiao Tung University (Taiwan), National Cheng Kung University (Taiwan)
Temporal focusing multiphoton excitation microscopy (TFMPEM) offers faster image acquisition compared to the traditional point-scanning multiphoton excitation microscopy (PSMPEM). Nevertheless, TFMPEM encounters challenges due to scattering in deeper biotissue layers. In this study, we introduce a multi-stage 3D U-Net approach to enhance TFMPEM image quality specifically for shallow layers without extending acquisition time. Furthermore, we propose a network based on ConvLSTM and PhyCell to predict images of deeper layers. This network learns from TFMPEM-restored images and enables precise analysis of deeper biological specimen layers within PSMPEM images.
 Preview abstract
|
|
|
16:15 - 16:30
|
Manuscript ID. 0089
Paper No. 2023-SAT-S0605-O002
|
| Yun-Chien Hung
|
Intelligent optical bone densitometry with U-Net segmentation for bone mineral density estimation
Yun-Chien Hung, National Yang Ming Chiao Tung University (Taiwan); Wei-Chun Chang, Taipei Municipal Wan fang Hospital (Taiwan); Tsai-Hsueh Leu, Taipei City Hospital Renai Branch (Taiwan); Yi-Min Wang, Gautam Takhellambam, Chia-Wei Sun, National Yang Ming Chiao Tung University (Taiwan)
In an aging society, early detection of osteoporosis is imperative due to the increased risk of bone injury associated with low bone mineral density (BMD). Our study introduces an intelligent optical bone densitometer (iOBD) combined with deep learning for estimating BMD in specific body regions. To accurately capture the wrist position, the target region for iOBD measurement, we utilize U-net for automated biomedical image segmentation to generate a mask for the wrist image. We can obtain noise-free images by multiplying the mask with original wrist images, facilitating subsequent deep learning analysis to estimate BMD in multiple body regions accurately.
Preview abstract
|
|
|
16:30 - 16:45
|
Manuscript ID. 0113
Paper No. 2023-SAT-S0605-O003
|
| Yi-Chong Wu
|
Liquid-crystal aptasensing of a single circulating tumor cell
Tsung-Keng Chang, Yi-Chong Wu, National Yang Ming Chiao Tung University (Taiwan); Mon-Juan Lee, Chang Jung Christian University (Taiwan); Wei Lee, National Yang Ming Chiao Tung University (Taiwan)
Circulating tumor cells (CTCs) are the key prognostic biomarker for evaluating the treatment response of cancer patients but their rarity in the bloodstream is the major challenge. We demonstrated a label-free liquid crystal (LC) cytosensor, by adopting an aptamer against epithelial cell adhesion molecule (EpCAM) to capture the EpCAM positive cancer cells. The optical biosensing approach resulted in a limit of detection (LOD) of 5 CTCs. Through the dielectric approach, we further improved the LOD to a single CTC. This study manifested the detection of single-cell CTCs in cancer cell-spiked human serum and whole blood using the LC-based dielectric cytosensor.
Preview abstract
|
|
|
16:45 - 17:00
|
Manuscript ID. 0308
Paper No. 2023-SAT-S0605-O004
|
| Hsiang-Fu Huang
|
Application of optical coherence tomography and deep learning for intraoperative lung cancer grading diagnosis
Hsiang-Fu Huang, National Yang Ming Chiao Tung University (Taiwan); Hung-Chang Liu, Mackay Memorial Hospital (Taiwan); Miao-Hui Lin, Rui-Cheng Zeng, Chia-Wei Sun, National Yang Ming Chiao Tung University (Taiwan)
This research proposes a novel human-machine interface (HMI) that automatically identifies types of lung lesions during surgery via combining mobile optical coherence tomography (OCT) and deep learning algorithms. With over 80 % sensitivity and specificity, this technique facilitates rapid histologically graded diagnosis, providing fast information to clinicians and offering a cost-effective approach for early detection and treatment guidance.
Preview abstract
|
|