AI4COVID-19: AI enabled preliminary diagnosis for COVID-19 from cough samples via an app

Abstract

Background: The inability to test at scale has become humanity's Achille's heel in the ongoing war against the COVID-19 pandemic. A scalable screening tool would be a game changer. Building on the prior work on cough-based diagnosis of respiratory diseases, we propose, develop and test an Artificial Intelligence (AI)-powered screening solution for COVID-19 infection that is deployable via a smartphone app. The app, named AI4COVID-19 records and sends three 3-s cough sounds to an AI engine running in the cloud, and returns a result within 2 min.

Methods: Cough is a symptom of over thirty non-COVID-19 related medical conditions. This makes the diagnosis of a COVID-19 infection by cough alone an extremely challenging multidisciplinary problem. We address this problem by investigating the distinctness of pathomorphological alterations in the respiratory system induced by COVID-19 infection when compared to other respiratory infections. To overcome the COVID-19 cough training data shortage we exploit transfer learning. To reduce the misdiagnosis risk stemming from the complex dimensionality of the problem, we leverage a multi-pronged mediator centered risk-averse AI architecture.

Results: Results show AI4COVID-19 can distinguish among COVID-19 coughs and several types of non-COVID-19 coughs. The accuracy is promising enough to encourage a large-scale collection of labeled cough data to gauge the generalization capability of AI4COVID-19. AI4COVID-19 is not a clinical grade testing tool. Instead, it offers a screening tool deployable anytime, anywhere, by anyone. It can also be a clinical decision assistance tool used to channel clinical-testing and treatment to those who need it the most, thereby saving more lives.

Keywords: Artificial intelligence; COVID-19; Pre-screening; Preliminary medical diagnosis; Public healthcare.

Fig. 1

Visualization of features for the four classes via t-SNE (gray triangles correspond to normal, blue circles correspond to bronchitis, black stars correspond to pertussis and orange diamonds represent COVID-19 cough. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 2

Proposed system architecture and flow diagram of AI4COVID-19, showing snapshot of Smartphone App at user front-end and back-end cloud AI-engine blocks consisting of Cough Detector block (further elaborated in Fig. 4 and Section 2.3) and COVID-19 diagnosis block containing Deep Transfer Learning-based Multi-Class classifier (DTL-MC), Classical Machine Learning-based Multi-Class classifier (CML-MC) and Deep Transfer Learning-based Binary-Class classifier (DTL-BC) (further elaborated in Fig. 5 and Section 2.3).

Fig. 3

A flow chart highlighting the steps of the proposed system.

Fig. 4

Cough detection classifier.

Fig. 5

Classical Machine Learning-based Multi-Class classifier (CML-MC).

Fig. 6

Normalized mean confusion matrix for cough detection (in percentage) using 5-fold cross validation.

Fig. 7

Mean model loss for 5-fold cross validation of cough detector.

Fig. 8

Normalized mean confusion matrix for cough diagnosis (in percentage) for DTL-MC using 5-fold cross validation.

Fig. 9

Mean model loss for 5-fold cross validation of DTL-MC.

Fig. 10

Normalized mean confusion matrix for cough diagnosis (in percentage) for CML-MC using 5-fold cross validation.

Fig. 11

Overall accuracy CDF for varying k-fold experiments in CML-MC approach.

Fig. 12

Normalized mean confusion matrix for cough diagnosis (in percentage) for DTL-BC using 5-fold cross validation.

Fig. 13

Mean model loss for 5-fold cross validation of DTL-BC.