Related Experiment Video
Updated: Oct 26, 2025
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
14.4K
Deep Learning-Based Approaches to Improve Classification Parameters for Diagnosing COVID-19 from CT Images
1Ministry of Health of Republic of Turkey, Ankara, Turkey.
Cognitive Computation
|July 26, 2021
Summary
This study introduces new pipeline methods to improve the accuracy of diagnosing COVID-19 pneumonia from CT scans. The proposed approaches significantly reduce misclassifications, enhancing diagnostic reliability for COVID-19 detection.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Pulmonology
Background:
- COVID-19 virus causes severe pneumonia with characteristic lung opacities on CT scans.
- Accurate diagnosis of COVID-19 pneumonia is crucial for patient management and preventing disease spread.
- Current diagnostic methods may face challenges with false-negative and false-positive results.
Purpose of the Study:
- To develop and evaluate novel pipeline approaches for improving the classification accuracy of COVID-19 diagnosis using CT lung images.
- To reduce false-negative, false-positive, and total misclassified images in differentiating COVID-19 from non-COVID-19 and COVID-19 pneumonia from other pneumonias.
- To assess the performance enhancement offered by these pipeline approaches in conjunction with a convolutional neural network (CNN).
Main Methods:
- Utilized a dataset of 4320 CT lung images for COVID-19/non-COVID-19 classification and 3801 images for COVID-19 pneumonia/other pneumonia classification.
- Employed a 24-layer convolutional neural network (CNN) architecture for image classification.
- Implemented five new pipeline approaches, incorporating local binary pattern (LBP) and dual-tree complex wavelet transform (DT-CWT) for feature extraction and image enhancement.
Main Results:
- Pipeline approaches improved COVID-19/non-COVID-19 classification metrics, with AUC reaching 0.9923 (vs. 0.9890 without pipelines).
- For COVID-19 pneumonia/other pneumonia classification, pipeline approaches yielded higher AUC of 0.9615 (vs. 0.9370 without pipelines).
- The proposed methods demonstrated increased sensitivity, specificity, and accuracy in both classification tasks.
Conclusions:
- The novel pipeline approaches significantly enhance the classification performance for diagnosing COVID-19 and COVID-19 pneumonia from CT images.
- These methods effectively reduce misclassification rates, leading to more reliable diagnostic outcomes.
- The study highlights the potential of optimized image processing pipelines to improve AI-driven medical image analysis for infectious diseases.
