# CoVid-19 Detection leveraging Vision Transformers and Explainable AI

**Authors:** Pangoth Santhosh Kumar, Kundrapu Supriya, Mallikharjuna Rao K  
**arXiv:** [2307.16033](https://arxiv.org/abs/2307.16033)  
**Date:** 2023-07-29  
**Summary page:** https://www.summarizepaper.com/en/paper/2307.16033v1/

## Key points

- Lung disease is a prevalent health issue worldwide, posing a significant risk to people's health and quality of life.
- It is responsible for five of the top thirty leading causes of death, including COVID-19, pneumonia, and tuberculosis.
- Early diagnosis of lung diseases is crucial for better patient outcomes and long-term survival.
- Various models utilizing machine learning and image processing techniques have been developed to aid in early detection.
- Deep learning algorithms offer promising potential for autonomous, rapid, and accurate identification of lung diseases based on medical imaging.
- Different deep learning strategies such as convolutional neural networks (CNN), vanilla neural networks, visual geometry group-based networks (VGG), and capsule networks have been employed to forecast lung diseases.
- Standard CNNs exhibit poor performance when dealing with rotated or tilted images.
- A vision transformer-based approach has been proposed as an end-to-end framework for diagnosing lung disorders.
- The architecture includes data augmentation, model training, and evaluation.
- A specialized Compact Convolution Transformer (CCT) model has been tested and evaluated on datasets like the Covid 19 Radiography Database to detect lung diseases such as pneumonia, COVID-19, and lung opacity.
- The CCT model has demonstrated improved accuracy in both training and validation on the Covid 19 Radiography Database compared to other models.
- This approach leverages vision transformers to enhance the performance of deep learning models in diagnosing lung diseases accurately.

## Summary

Lung disease is a prevalent health issue worldwide, posing a significant risk to people's health and quality of life. It is responsible for five of the top thirty leading causes of death, including COVID-19, pneumonia, and tuberculosis. Early diagnosis of lung diseases is crucial for better patient outcomes and long-term survival. Various models utilizing machine learning and image processing techniques have been developed to aid in early detection. Deep learning algorithms offer promising potential for autonomous, rapid, and accurate identification of lung diseases based on medical imaging. Different deep learning strategies such as convolutional neural networks (CNN), vanilla neural networks, visual geometry group-based networks (VGG), and capsule networks have been employed to forecast lung diseases. However, standard CNNs exhibit poor performance when dealing with rotated or tilted images. To address this limitation, a vision transformer-based approach has been proposed in this study as an end-to-end framework for diagnosing lung disorders. The architecture includes data augmentation, model training, and evaluation. A specialized Compact Convolution Transformer (CCT) model has been tested and evaluated on datasets like the Covid 19 Radiography Database to detect lung diseases such as pneumonia, COVID-19, and lung opacity. The CCT model has demonstrated improved accuracy in both training and validation on the Covid 19 Radiography Database compared to other models. This approach leverages vision transformers to enhance the performance of deep learning models in diagnosing lung diseases accurately. In addition to the existing summary's authors Pangoth Santhosh Kumar, Kundrapu Supriya, Mallikharjuna Rao K from IIIT Naya Raipur are mentioned as contributors in the field of Data Science and Artificial Intelligence.