Scat-NET: COVID-19 diagnosis with a CNN model using scattergram images

Abstract

The acute respiratory syndrome COVID-19 disease, which is caused by SARS-CoV-2, has infected many people over a short time and caused the death of more than 2 million people. The gold standard in detecting COVID-19 is to apply the reverse transcription polymerase chain reaction (RT-PCR) test. This test has low sensitivity and produces false results of approximately 15%–20%. Computer tomography (CT) images were checked as a result of suspicious RT-PCR tests. If the virus is not infected in the lung, the virus is not observed on CT lung images. To overcome this problem, we propose a 25-depth convolutional neural network (CNN) model that uses scattergram images, which we call Scat-NET. Scattergram images are frequently used to reveal the numbers of neutrophils, eosinophils, basophils, lymphocytes and monocytes, which are measurements used in evaluating disease symptoms, and the relationships between them. To the best of our knowledge, using the CNN together with scattergram images in the detection of COVID-19 is the first study on this subject. Scattergram images obtained from 335 patients in total were classified using the Scat-NET architecture. The overall accuracy was 92.4%. The most striking finding in the results obtained was that COVID-19 patients with negative RT-PCR tests but positive CT test results were positive. As a result, we emphasize that the Scat-NET model will be an alternative to CT scans and could be applied as a secondary test for patients with negative RT-PCR tests.

Assessing population-sampling strategies for reducing the COVID-19 incidence

“…One of the drawbacks of RT-PCR testing is that produces false results in a significant percentage of the cases. In order to increase the diagnosis accuracy, complementary techniques [ 35 ] can be used in combination with sampling. Due to the economic cost of the testing campaigns, other alternatives have been explored to reduce these costs with minimum detection loss of positive cases.…”

A convolutional neural network‐based, quantitative complete blood count scattergram‐mapping framework promptly screens acute promyelocytic leukemia with high sensitivity

“…[12][13][14][15][16][17][18] In addition, there have been studies applying CNN to CBC scattergrams for other classification tasks. [19][20][21] Specifically, deep learning models have been applied to predict APL from cell images. 17,22 However, patients who have potential APL and normal CBC counts would not be flagged and potentially would be missed under the current review rules recommended by the International Society of Laboratory Hematology (ISLH).…”

“…Recently, progress has been made in developing convolutional neural networks (CNNs) to recognize cell subtypes from microscopic images of peripheral blood or bone marrow smears 12–18 . In addition, there have been studies applying CNN to CBC scattergrams for other classification tasks 19–21 . Specifically, deep learning models have been applied to predict APL from cell images 17,22 .…”