Detection of COVID-19 from CT Lung Scans Using Transfer Learning

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Disease Diagnostics

Background:

• Accurate and timely detection of COVID-19 is crucial for patient management and public health.
• Computed Tomography (CT) scans are widely used for diagnosing COVID-19, but interpretation can be challenging.
• Traditional diagnostic methods may be time-consuming or lack sufficient accuracy.

Purpose of the Study:

• To investigate the efficacy of transfer learning architectures for COVID-19 detection using CT lung scans.
• To evaluate the performance of different transfer learning models and image preprocessing techniques.
• To identify optimal models for automated COVID-19 diagnosis from medical imaging.

Main Methods:

• Utilized various transfer learning architectures for image classification.
• Applied Histogram Equalization and Contrast Limited Adaptive Histogram Equalization for image enhancement.
• Trained and validated models on a dataset of COVID-19 positive and negative CT lung scans.
• Assessed model performance using metrics such as accuracy and recall.

Main Results:

• Transfer learning frameworks demonstrated effectiveness in detecting COVID-19 from CT scans.
• The VGG-19 model, combined with Contrast Limited Adaptive Histogram Equalization, yielded the best performance.
• The top-performing model achieved an accuracy of 95.75% and a recall of 97.13% on the SARS-CoV-2 dataset.

Conclusions:

• Transfer learning offers a viable alternative to existing methods for COVID-19 detection in CT imaging.
• Image preprocessing techniques significantly impact the performance of deep learning models in this context.
• The VGG-19 architecture with CLAHE shows strong potential for clinical application in COVID-19 diagnosis.