June 2021
Spotlight Summary by Pascal Picart
Digital holographic deep learning of red blood cells for field-portable, rapid COVID-19 screening
Digital holographic microscopy has proven to be very powerful for quantitative evaluation of biological samples; especially because it provides measurements of the optical path difference (OPD). The use of such OPD from biological samples can be of great benefit when data are associated to deep learning approaches. The paper by T. O’Connor et al. proposes a low cost and rapid method for the screening of red blood cells infected by COVID-19. Their clever system is based on digital shearing microscopy built with 3D printing additive manufacturing, making it portable and cost-efficient. The authors propose deep learning processing (with a bi-directional long short-term memory neural network) of the OPD temporal fluctuation of individual red blood cells as a discriminator for classification. Video holograms of red blood cells were recorded on-site using their compact microscope, then cells were individually segmented, and features were extracted at each timeframe of the video data. Analysis of 1474 red blood cells from 10 COVID-19 positive individuals and 14 healthy healthcare workers was carried out. They demonstrate that their system correctly identified the COVID status of 87.50% of all individuals (21/24), 8/10 identified as COVID positive, and 13/14 identified as healthy. Although requiring further investigations to overcome limitations in the selection of infected and non-infected individuals, the paper provides strong evidence of differences in red blood cell behavior between the two studied populations and the ability to classify individuals based on red blood cell examination.
You must log in to add comments.
Add Comment
You must log in to add comments.
Article Information
Digital holographic deep learning of red blood cells for field-portable, rapid COVID-19 screening
Timothy O’Connor, Jian-Bing Shen, Bruce T. Liang, and Bahram Javidi
Opt. Lett. 46(10) 2344-2347 (2021) View: Abstract | HTML | PDF