Identification of Human Papillomavirus from Super-Resolution Microscopic Images Generated Using Deep Learning Architectures

Abstract

This chapter presents five deep learning architectures for identification of Human papillomavirus (HPV) through generation of super-resolution (SR) images by fourfolds. Specifically, generative adversarial deep learning networks (GAN) and a texture-based vision transformer (TTSR) architecture are applied and evaluated. As such, the generated SR images are able to display the same way a high-resolution image offers in identification of HPV-like structures. In comparison, TTSR appears to perform the best with PSNR and SSIM being 28.70 and 0.8778, respectively, whereas 25.80/0.7910, 18.35/0.5059. 30.31/0.8013 and 28.07/0.6074 are observed for the methods of RCAN, Pix2Pix, CycleGAN and ESRGAN, respectively. With regard to sensitivity and specificity when detecting HPV clusters, TTSR also leads with 83.6% and 83.33%, respectively. It appears the computational SR images are capable to differentiate distinguishing features of HPV-like particles and to determine the effectiveness of anti-HPV agents, holding promise providing insights into the formation stage of a cancer from HPV in the near future. Latest version of May 2022.

Appendices

Appendix A. Sample Preparation for Confocal Fluorescent Microscopy

The cervical cancer cell lines, CaSki containing HPV16 sequences and C33a without HPV, were used in this study. Both cell lines were grown on cover slips in a six-well culture plate containing RPMI culture media with 10% Fetal calf serum and 1% penicillin/streptomycin. They were kept in a humidified incubator with 95% air and 5% carbon dioxide for 2 days until the cells reached 70–80% confluence. The cells were washed by PBS three times with 1 min each time before they were fixed by 4% paraformaldehyde for 10 min. They were then exposed to 0.1% Triton-100 for 5 min following PBS washes. 50% house serum was then added for 8 min, then it was removed. Next, 200ul 1 in 100 dilution of anti-mouse HPV E6/E7 antibody (Abcam, UK) was added in and the cells were left at the room temperature for 2 h before they were washed again and 100ul biotinylated second antibody (ABC Universal Kit, Vector lab, UK) was added. After 30 min, the cells were washed again and then tertiary antibody was added and left for 20 min. Finally, 100 ul TSA/FITC amplification reagent exposed to the cells for 6 min in the darkness before the cells were washed. DAPI containing mounting media was added on the labelled slides and cells attached on the cover slips were sealed inside for microscopic viewing.

Appendix B. Sample Preparation for TEM Scanning

Cells were grown and prepared in a similar way as described above in Section S1 until the procedure reached at fixation step. Instead of fixing cells by paraformaldehyde, 2.5% glutaraldehyde in PBS buffer at pH 7 was used and the cells were fixed for 3 h at RT before further steps taking place.

In addition, before the scanning, these samples undertake a series of standard preparation processes, including (1) flat embedding into EM capsules and polymerize for 6 h at 16 °C, (2) sectioned to 70 nm thick using a Leica UC7 ultramicrotome, (3) collected onto copper mesh grids and (4) stained in lead citrate for 5 min.