An Approach for Identifying COVID-19 from CT-SCANS using Machine Learning Algorithms

Abstract

This project introduces a machine learning approach for rapid COVID-19 detection using chest CT scans, addressing delays in RT-PCR testing. After enhancing and reducing the dimensionality of CT images, models like KNN, Decision Tree, SVM, and Random Forest were evaluated. The best—Random Forest, KNN, and SVM—were combined into a soft voting ensemble for improved accuracy. A Streamlit web app was developed to allow real-time image upload and instant COVID vs non-COVID predictions, providing a scalable diagnostic aid.

Introduction

The COVID-19 pandemic highlighted the need for rapid, accurate diagnosis to control virus spread. While RT-PCR is the standard test, it has limitations in sensitivity and speed. Medical imaging, especially CT scans, offers a valuable alternative due to higher sensitivity in detecting lung abnormalities related to COVID-19. However, interpreting CT scans requires expert radiologists, creating a bottleneck.

To address this, the project proposes an automated machine learning system that classifies CT scan images as COVID or Non-COVID, aiming to support healthcare professionals with a fast, reliable, and scalable diagnostic tool.

Objectives and Approach:

• Enhance CT images and reduce feature dimensions using Principal Component Analysis (PCA).

• Train multiple classifiers (Random Forest, KNN, SVM, etc.) and combine the best three via a soft voting ensemble to improve accuracy.

• Deploy the model in a Streamlit web application enabling real-time diagnosis through user-friendly CT image upload.

Literature Review:

• Prior studies have applied traditional and deep learning techniques on chest X-rays and CT images, achieving varying accuracies (75% to over 99%) using CNNs, ensemble methods, and feature selection.

• Challenges include small datasets, lack of standardization, and limited clinical deployment.

Methodology:

• Preprocess images by converting to grayscale, resizing, normalizing, and enhancing contrast.

• Extract features and apply PCA to retain 90% variance for efficient modeling.

• Train and evaluate multiple machine learning models individually.

• Use soft voting ensemble to combine top-performing models (Random Forest, KNN, SVM).

• Deploy the saved ensemble model on a Streamlit app for instant COVID/Non-COVID classification.

Implementation:

• Developed in Python using libraries like Scikit-learn, OpenCV, and Streamlit.

• Ensemble method averages prediction probabilities from individual classifiers for robust final decision.

• Web app enables easy CT scan uploads and instant diagnostic feedback.

Results:

• The system provides a practical, real-time tool for COVID-19 detection from CT scans, improving diagnostic speed and aiding medical professionals.

Conclusion

This project presents an effective machine learning-based system for detecting COVID-19 from chest CT scan images, achieving a high accuracy of 99.84%. It combines image preprocessing, histogram equalization, PCA-based dimensionality reduction, and multiple classifiers—KNN, SVM, and Random Forest—into a soft voting ensemble for robust performance. The system is deployed through a Streamlit web interface, enabling real-time image classification. Evaluation results, including a near-perfect confusion matrix, demonstrate the model’s reliability. This approach offers a fast, automated alternative to RT-PCR testing and holds promise for future integration with larger datasets and advanced deep learning techniques.

References

[1] Videla, Lakshmi Sarvani, et al. \"Convolution Neural Networks based COVID-19 Detection using X-ray Images of Human Chest.\" 2022 8th International Conference on Smart Structures and Systems (ICSSS). IEEE, 2022. [2] Sayed, Safynaz Abdel-Fattah, Abeer Mohamed Elkorany, and Sabah Sayed Mohammad. \"Applying different machine learning techniques for prediction of COVID-19 severity.\" Ieee Access 9 (2021): 135697-135707. [3] Tang, Shanjiang, et al. \"EDL-COVID: Ensemble deep learning for COVID-19 case detection from chest X-ray images.\" IEEE Transactions on Industrial Informatics 17.9 (2021): 6539-6549. [4] Haq, Amin Ul, et al. \"Deep Learning Approach for COVID-19 Identification.\" 2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP). IEEE, 2021. [5] Irmak, Emrah. \"A novel deep convolutional neural network model for COVID-19 disease detection.\" 2020 Medical Technologies Congress (TIPTEKNO). IEEE, 2020.

Copyright

Copyright © 2025 Joshna Gorantala, G. Narasimham. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.