LHSPred - Lung Health Severity Prediction

A web based tool to predict lung health severity in COVID-19 patients.

It uses Support Vector Regressor (SVR) and Multi-layer Perceptron Regressor (MLPR) trained with COVID-19 patients' data to determine a score (CT severity score) that evaluates the involvement of lesions in the lungs. This computed score is then used to predict risk of pneumonia.

Cite as:

Bhattacharjee, S., Saha, B., Bhattacharyya, P., & Saha, S. (2022). LHSPred: A web based application for predicting lung health severity. Biomedical signal processing and control, 77, 103745. https://doi.org/10.1016/j.bspc.2022.103745.

Using the tool

LHSPred is available at: http://dibresources.jcbose.ac.in/ssaha4/lhspred.

To know more about the methodology, please refer to the About page.

The dataset used by the regression models is available here. The patient data was originally published by Feng, Z. et al. (2020).

Development

It is deployed in a Apache HTTPD server. Python libraries used :

• numpy
• scikit-learn (Version-0.20.0)
• joblib (Version-0.14.1)
• scipy (Version-1.4.1 for Python3 and version-1.2.3 for Python2)
• pathlib
• statistics

Currently, trained models of only scikit-learn version-0.20.0 are saved with both Python2 and Python3. Plotly JS library is used for density plot in the prediction output.

Team

• Sudipto Bhattacharjee ([email protected])
  Ph.D. Scholar,
  Department of Computer Science and Engineering,
  University of Calcutta, Kolkata, India.
  
• Dr. Banani Saha ([email protected])
  Associate Professor,
  Department of Computer Science and Engineering,
  University of Calcutta, Kolkata, India.
• Dr. Parthasarathi Bhattacharyya ([email protected])
  Consultant Pulmologist,
  Institute of Pulmocare and Research,
  Kolkata, India.
• Dr. Sudipto Saha ([email protected])
  Associate Professor,
  Division of Bioinformatics,
  Bose Institute, Kolkata, India.

Please contact Dr. Sudipto Saha regarding any further queries.