[Paper Review] TeliNet, a simple and shallow Convolution Neural Network (CNN) to Classify CT Scans of COVID-19 patients.

Mohammad Nayeem Teli|arXiv (Cornell University)|Jul 10, 2021

COVID-19 diagnosis using AI18 references1 citations

TL;DR

TeliNet is a lightweight, shallow Convolutional Neural Network designed to classify CT scans of COVID-19 patients with high accuracy. It achieves superior F1 scores compared to VGGNet and benchmark models while maintaining reduced computational complexity.

ABSTRACT

Hundreds of millions of cases and millions of deaths have occurred worldwide due to COVID-19. 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 diagnosis of the disease and isolation of the patients to prevent any spreads play a huge role. Machine Learning approaches have assisted the diagnosis of COVID-19 cases by analyzing chest X-ray and CT-scan images of patients. In this research we present a simple and shallow Convolutional Neural Network based approach, TeliNet, to classify CT-scan images of COVID-19 patients. Our results outperform the F1 score of VGGNet and the benchmark approaches. Our proposed solution is also more lightweight in comparison to the other methods.

Motivation & Objective

• To develop a lightweight deep learning model for rapid and accurate classification of COVID-19 from CT scans.
• To address the need for efficient diagnostic tools in regions with limited computational resources.
• To improve classification performance over existing benchmark models like VGGNet using a simpler architecture.
• To reduce model complexity without sacrificing diagnostic accuracy in COVID-19 detection.

Proposed method

• Designing a shallow Convolutional Neural Network (CNN) architecture with fewer layers and parameters to reduce computational load.
• Utilizing convolutional layers to extract spatial features from CT scan images.
• Applying standard deep learning components such as ReLU activation and pooling layers for feature learning and dimensionality reduction.
• Training the model end-to-end on a dataset of labeled CT scans to optimize classification performance.
• Employing standard optimization techniques such as stochastic gradient descent for model training.
• Evaluating performance using standard metrics including F1 score, precision, and recall on a test set of CT scans.

Experimental results

Research questions

• RQ1 Can a shallow and lightweight CNN achieve better F1 scores than deeper models like VGGNet in classifying COVID-19 from CT scans?
• RQ2 To what extent does model simplicity reduce computational cost while maintaining high diagnostic accuracy?
• RQ3 How does TeliNet compare to existing benchmark models in terms of F1 score and inference efficiency?
• RQ4 Is a minimal architecture sufficient to extract discriminative features from CT scan images for reliable COVID-19 classification?

Key findings

• TeliNet achieves a higher F1 score than VGGNet on the same CT scan classification task.
• The proposed model outperforms benchmark approaches in terms of classification performance.
• TeliNet maintains a lightweight architecture, making it suitable for deployment in resource-constrained environments.
• The model demonstrates strong diagnostic performance despite its simplicity and reduced parameter count.