COVID-19 Infection and Machine Learning Using Artificial Intelligence (AI)

East Suffolk and North Essex NHS Foundation Trust215 enrolled

Overview

COVID-19 infection is currently confirmed by a complex, multiple-step procedure starting with a mucosal swab, followed by viral RNA extraction and processing and qPCR. This study aims to explore a novel method using machine learning and artificial intelligence (AI) algorithm to diagnose COVID-19 infection through the morphological analysis of lymphocyte subset in the peripheral blood. This study will also risk stratify patients with COVID 19 infection based on the above finding along with other clinical, haematological and biochemical parameters with a view to predict clinical outcome with high sensitivity and specificity.

This is an observational study which will be carried out at East Suffolk and North Essex NHS Foundation Trust (ESNEFT) in collaboration with University of Suffolk (UoS). Investigators aim to analyse subsets of lymphocytes in the prospective blood smear slides using machine learning and AI algorithm obtained from participants with a positive qPCR test for COVID-19 who have required a hospital admission. The control group will consist of archived blood smear slide data from patients both with i) non-suspected viral infections, and ii) those with a non-COVID-19 viral infection obtained prior to the emergence of COVID-19 infection in the United Kingdom. In total, 785 blood smear slides will be analysed. The aim of this study is to establish the diagnosis of COVID 19 infection based on lymphocyte morphology on patients with COVID-19 infection from other patients with non COVID -19 viral infections. A high definition single cell lymphocyte image from patients with COVID 19 infection and control group will be analysed using open source histopathology imaging software CellProfiler against very fine cytoplasmic and nuclear details of the cells through supervised and unsupervised machine learning algorithm to identify recurring pattern that is unique to COVID 19 infection. The study will also assess other relevant clinical, haematological and biochemical parameters in conjunction with the above morphological features to develop a risk stratification tool to predict the clinical outcome of patients with COVID-19 infection with high specificity and sensitivity using bioinformatics pipeline.

Study Type

OBSERVATIONAL

Enrollment

215

Conditions

Eligibility

Sex: ALLMin age: 18 Years

Medical Language ↔ Plain English

Inclusion Criteria: * Female or male participants * Aged over 18 years old (no upper age limit) * Patients with SARS-COV-2 positive diagnosis based on qPCR (Study COVID 19 group) * Peripheral blood smear slides from patients with no viral infection, reposited in the laboratory slides archive within the facility prior to the emergence of COVID-19 infection in the United Kingdom (Control group) * Peripheral blood smear slides from patients with a non-SARS-CoV-2 viral infection that were reposited in the laboratory slides archive within the facility prior to the emergence of COVID-19 infection in the United Kingdom (Control group). Exclusion Criteria: * Patients that are less than 18 years old * Patients with SARS-COV-2 negative diagnosis based on qPCRPatients who have been haematological malignancies with lymphocytosis as predominant manifestation. * Patients who have lymphopenia in the past due to underlying inflammatory disorders. * Patients who have lymphopenia due to previous cytotoxic or immunosuppressive therapy. * Positive diagnosis of Human Immunodeficiency Virus (HIV).

Outcomes

Primary Outcomes

Diagnosis of COVID-19.

Determine whether lymphocytes alone can diagnose COVID-19 disease with high specificity and sensitivity, using AI-based image analytical modelling.

Time frame: 6 months

Secondary Outcomes

Severity of COVID-19 infection modelling

The secondary outcome measure of the study will be to create risk stratification modelling, to aid in predicting the severity and mortality of the infection, based on our above-mentioned, novel diagnostic tool and additional clinical, haematological and biochemical parameters; ensuring high specificity, with consequent facilitated management of patients both in a hospital and outpatient setting. The model proposed intends to use and evaluate the clinical parameters including oxygen saturation at the time of venesection, and other vital statistics, including: pulse, blood pressure and respiratory rate, along with other parameters such as LDH, ferritin, C-reactive protein (CRP), D-dimers, renal function, all together helping to predict disease outcome and severity.