LHSPred: A web based application for predicting lung health severity

Sudipto Bhattacharjee; Banani Saha; Parthasarathi Bhattacharyya; Sudipto Saha

doi:10.1016/j.bspc.2022.103745

LHSPred: A web based application for predicting lung health severity

Biomed Signal Process Control. 2022 Aug:77:103745. doi: 10.1016/j.bspc.2022.103745. Epub 2022 May 12.

Authors

Sudipto Bhattacharjee¹, Banani Saha¹, Parthasarathi Bhattacharyya², Sudipto Saha³

Affiliations

¹ Department of Computer Science and Engineering, University of Calcutta, JD-2, Sector-III, Salt Lake, Kolkata 700098, India.
² Institute of Pulmocare and Research, DG-8, Action Area-1, New Town, Kolkata 700156, India.
³ Division of Bioinformatics, Bose Institute, P-1/12 C.I.T. Scheme VII-M, Kolkata 700054, India.

Abstract

Background and objectives: The computed tomography (CT) scan facilities are crucial for diagnosis of pulmonary diseases and are overburdened during the current pandemic of novel coronavirus disease 2019 (COVID-19). LHSPred (Lung Health Severity Prediction) is a web based tool that enables users to determine a score that evaluates CT scans, without radiologist intervention, and predict risk of pneumonia with features of blood examination and age of patient. It can help in early assessment of lung health severity of patients without CT-scan results and also enable monitoring of post-COVID lung health for recovered patients.

Methods: This tool uses Support Vector Regression (SVR) and Multi-Layer Perceptron Regression (MLPR), trained on COVID-19 patient data reported in the literature. It allows to compute a score (CT severity score) that evaluates the involvement of lesions in lung lobes and to predict risk of pneumonia. A web application was implemented that uses the trained regression models.

Results: The application has proven to be effective and user friendly in a clinical setting for pulmonary disease treatment. The SVR model achieved Pearson correlation coefficient (PCC) of 0.77 and mean absolute error (MAE) of 2.239 while determining the computed tomography (CT) severity score. The MLPR model achieved PCC of 0.77 and MAE of 2.309. Thus, it can be applied as a useful tool in predicting pneumonia in the post COVID-19 era.

Conclusion: LHSPred can be used as a decision support system by the clinicians and as a tool for self-assessment by the patients with only six blood test input features.

Keywords: COVID-19; HRCT scan image; Lung health; Multi-layer Perceptron Regression; Pneumonia; Support Vector Regression.