Automatic segmentation of COVID-19 from computed tomography images using modified U-Net model-based majority voting...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Coronavirus disease (COVID-19) presents a significant global health challenge.
• Accurate and timely detection of COVID-19 is crucial for patient management and public health.
• Computed tomography (CT) imaging is a key diagnostic tool for COVID-19.

Purpose of the Study:

• To develop and evaluate a deep learning model for automated COVID-19 detection using CT images.
• To improve the efficiency and accuracy of COVID-19 diagnosis through advanced image segmentation techniques.

Main Methods:

• A novel deep learning model was proposed, based on modifying the U-Net segmentation architecture.
• The encoder of the U-Net model was enhanced using various deep learning architectures: VGG16, ResNet101, DenseNet121, InceptionV3, and EfficientNetB5.
• A majority voting principle was employed to combine the outputs from the modified U-Net models for a final prediction.

Main Results:

• The proposed model achieved a Dice score of 85.03% for COVID-19 segmentation.
• The model demonstrated high diagnostic performance with 89.13% sensitivity and 99.38% specificity.
• Experimental results indicate the model's effectiveness on a dedicated COVID-19 segmentation test dataset.

Conclusions:

• The developed deep learning approach shows significant promise for accurate COVID-19 detection from CT images.
• This automated method can potentially reduce diagnostic time and costs for clinicians.
• The findings suggest a valuable tool to support clinical decision-making in the context of the COVID-19 pandemic.