Deep learning based fusion model for COVID-19 diagnosis and classification using computed tomography images.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• The COVID-19 pandemic necessitates rapid and accurate diagnostic tools.
• Limitations in rapid testing kits highlight the need for automated image-based diagnosis.
• Existing AI models for COVID-19 diagnosis primarily utilize X-ray images.

Purpose of the Study:

• To develop a novel Deep Learning Based MultiModal Fusion (DLMMF) technique for COVID-19 diagnosis and classification.
• To leverage Computed Tomography (CT) images for enhanced diagnostic capabilities.
• To improve automated detection of COVID-19 using advanced AI methods.

Main Methods:

• The DLMMF model employs Weiner Filtering (WF) for pre-processing CT images.
• Deep features are extracted and fused using VGG16 and Inception v4 models.
• A Gaussian Naïve Bayes (GNB) classifier is utilized for final classification of CT images.

Main Results:

• The DLMMF model demonstrated superior performance on the COVID-CT dataset (760 images).
• Achieved a maximum sensitivity of 96.53% and specificity of 95.81%.
• Attained an overall accuracy of 96.81% and an F-score of 96.73%.

Conclusions:

• The proposed DLMMF technique shows significant potential for accurate COVID-19 diagnosis from CT images.
• The multimodal fusion approach enhances diagnostic performance.
• This automated model can aid in efficient and reliable COVID-19 detection.