COVID-19 infection and machine learning using Artificial Intelligence [COVID-19]

  • Research type

    Research Study

  • Full title

    Rapid diagnosis of COVID-19 positive patients with Artificial Intelligence (AI) algorithm using clinical and image analytical parameters to evaluate the lymphocyte subsets in the peripheral blood.

  • IRAS ID

    282667

  • Contact name

    Mahesh Prahladan

  • Contact email

    mahesh.prahladan@esneft.nhs.uk

  • Sponsor organisation

    East Suffolk and North Essex Foundation NHS Trust

  • Duration of Study in the UK

    0 years, 6 months, 1 days

  • Research summary

    There is increasing evidence of numbers of specific white blood cells, namely the lymphocytes, being negatively affected in patients suffering from SARS-CoV-2 infection, causing COVID-19. The aim of our study is to establish whether the morphology of lymphocytes from COVID-19 patients is significantly different from that of control patients, i.e. with other unrelated viral infections or non suspected viral infections, to identify a unique disease “fingerprint“, based on a simple blood smear. \n\nInitially, individual control and COVID-19 positive lymphocyte images, obtained from peripheral blood smears, will be analysed in a double-blind study using open source histopathology imaging software CellProfiler against several recognised categories, such as nucleus morphology, cytoplasm-to-nucleus ratio, etc… Each category will be stratified in order to design a grading system that could potentially be correlated with other current relevant clinical, haematological and biochemical parameters, such as: pulse, blood pressure and respiratory rate, lactate dehydrogenase (LDH) ferritin, C-reactive protein (CRP) and renal function. \n\nFurthermore, upon establishment of a robust system of graded categories, we will develop an Artificial Intelligence (AI)-based automated processing workflow utilising or adapting open-source software to facilitate an end-to-end image analysis protocol. This will result in a cost-effective, reliable methodology capable of providing a rapid diagnostic tool with high sensitivity and specificity for COVID-19 detection. \nThis tool may ultimately help in the prediction of the clinical outcomes (severity and mortality) at the time of patient presentation, which potentially may help in the management both in a hospital and outpatient settings.\n\nWe predict that this innovative approach would limit, if not eliminate, the need for complex testing based upon the use of expensive, limited-in-stock reagents, and equipment. We envisage that this workflow will be adaptable to point of care diagnostic solutions, with a great advantage for both advanced and less sophisticated health systems. \n\n\n

  • REC name

    South West - Central Bristol Research Ethics Committee

  • REC reference

    20/SW/0079

  • Date of REC Opinion

    24 Apr 2020

  • REC opinion

    Favourable Opinion