Combining artificial neural networks and blood tests to diagnose COVID-19 infection

ABSTRACT

Background: Fast and accurate diagnosis of COVID-19 is important to prevent dissemination and disease progression. Artificial Intelligence is known as a universal fitting tool and can be used on the formulation of predictive models for the disease's diagnosis. Aims: Obtain a neural network (ANN) to diagnose patients as positive or negative COVID-19 based on patient data and blood tests. Methods: Data from 678 patients with moderate symptoms from the Anhembi Field Municipal Hospital (São Paulo-Brazil), followed between June/2020 and October/2020 were used. Covid-19 by RTPCR was confirmed in 460 patients. The inputs considered were sex, age, ethinicity, body mass index, tabagism, ex-tabagism, alveolar infiltrate, arterial hypertension, diabetes, heart rate, respiration rate, body temperature, oxygen saturation, D-dimer, activated partial thromboplastin time, prothrombin time, levels of hemoglobin, platelet, leukocytes, lymphocytes, monocytes, neutrophils, lactate dehydrogenase, C-reactive protein, and creatinine. Blood was collected at the patient's admission. The ANNs had 25 inputs and the output was the Covid-19 diagnosis. The best ANN was defined by a 5-fold cross-validation scheme. Then, a test step was performed to assess the model's performance. ANNs with one and two hidden layers were proposed. The number of neurons ranged from 5 to 35. Results: The best result was obtained with an ANN containing 25 and 30 neurons in the first and second hidden layers, respectively. All the statistical parameters found for the best model are shown at Table 1. The model presented accuracy of 83.3 %, high capacity for the prediction of true positives (PPV = 0.917 and LR+ = 5.188), and moderate probability to indicate false negatives (LR-= 0.202). Conclusions: The results showed that the ANNs are promising to diagnose Covid-19 based on clinical parameters and blood tests. After future refinements and proper validation, this model could be used to diagnose Covid-19 on daily basis.