TeliNet: Classifying CT scan images for COVID-19 diagnosis

TL;DR

TeliNet is a lightweight convolutional neural network that effectively classifies COVID-19 CT scans, outperforming benchmark and VGGNet models in a competitive challenge.

Contribution

The paper introduces TeliNet, a simple and shallow CNN architecture that improves COVID-19 CT scan classification accuracy while being more efficient than existing models.

Findings

TeliNet outperforms the competition benchmark in F1 macro score.

TeliNet surpasses VGGNet in classification performance.

The approach is more lightweight and computationally efficient.

Abstract

COVID-19 has led to hundreds of millions of cases and millions of deaths worldwide since its onset. The fight against this pandemic is on-going on multiple fronts. While vaccinations are picking up speed, there are still billions of unvaccinated people. In this fight against the virus, diagnosis of the disease and isolation of the patients to prevent any spread play a huge role. Machine Learning approaches have assisted in the diagnosis of COVID-19 cases by analyzing chest X-rays and CT-scan images of patients. To push algorithm development and research in this direction of radiological diagnosis, a challenge to classify CT-scan series was organized in conjunction with ICCV, 2021. In this research we present a simple and shallow Convolutional Neural Network based approach, TeliNet, to classify these CT-scan images of COVID-19 patients presented as part of this competition. Our results outperform the F1 `macro' score of the competition benchmark and VGGNet approaches. Our proposed solution is also more lightweight in comparison to the other methods.