UROP Proceedings 2020-21

School of Engineering Department of Chemical and Biological Engineering 87 Analyzing the Possible Water-Soluble Proteins for Deep-Tissue and Cancer-Targeting in Vivo Imaging Using a Photoacoustic Imaging System Supervisor: WONG Tsz Wai / CBE Student: LAI Ka Wai / SENG Course: UROP1100, Summer Early diagnosis of cancer can greatly increase the chance of curing the disease. Existing imaging modalities for cancer detection may be too costly and time-consuming. They also have disadvantages like the low molecular specificity and penetration depth. A photoacoustic imaging system can overcome the limitations and improve image quality. This study will investigate some possible contrast agents. The agent used should be biodegradable, cancer-targeting, and having a versatile molecular design, a near-infra-red absorption peak. The contrast agent can also potentially perform photodynamic therapy which makes the imaging system a cancer-curing device simultaneously. The system should apply single-wavelength differential imaging to improve the image quality. An optimal candidate for the protein in the contrast agent is BphP1. Development of a Deep Learning Approach for Transferring Gray-Scale Images Acquired from Photoacoustic Imaging to Histological-Stained Images Supervisor: WONG Tsz Wai / CBE Student: SHI Naiyu / BIEN Course: UROP1100, Fall Histology staining has been universally used in hospitals for pathologists to give diagnoses. However, the process of histology staining is complicated and takes a long time before the skilled pathologists could make a judgement. In this work, we explore the use of deep learning methods to transfer the gray-scale images obtained by microscopic imaging technique into histological-stained images. Deep learning, as a subfield of machine learning methods, is based on algorithms inspired by the structure and function of the brain, named artificial neural networks. We specifically look into a model of Generative Adversarial Network (GAN) called CycleGAN, study the code of it and also its capabilities. Development of a Deep Learning Approach for Transferring Gray-Scale Images Acquired from Photoacoustic Imaging to Histological-Stained Images Supervisor: WONG Tsz Wai / CBE Student: LO Shin Dawn / BIEN Course: UROP1100, Spring Surgery is often used to treat local-stage breast cancer (American Cancer Society, 2020; Wong, et al., 2017; Wong, et al., 2018); the current industry practise for evaluating the success of a surgery is to stain tissues post-surgery. In the case that a tumour is not completely removed, patients need to undergo follow-up surgeries, which is timely and costly. Photoacoustic microscopy is a promising imaging technique that can identify cancerous cells in vivo during surgery, reducing the need for follow-up surgeries. To improve the usability of PAT, deep learning is applied to translate PAT images into histological-like images that surgeons are accustomed to viewing. The following paper evaluates the use of deep learning for image translation and suggests other computational methods that could improve the functionality of PAT.

RkJQdWJsaXNoZXIy NDk5Njg=