COVID-19 Image Segmentation Based on Deep Learning and Ensemble Learning.

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Infectious Disease Diagnostics

Background:

• Medical imaging, particularly CT scans, is crucial for COVID-19 diagnosis and monitoring.
• Accurate segmentation of infected lung regions is essential for quantitative analysis and treatment assessment.
• Deep learning models offer promising avenues for automating medical image analysis.

Purpose of the Study:

• To develop an automated pipeline for segmenting COVID-19 infection areas in CT scans.
• To evaluate the performance of ensemble learning techniques (Bagging and Augmenting) on segmentation accuracy.
• To determine the most effective ensemble method for COVID-19 CT image segmentation.

Main Methods:

• Implementation of a deep convolutional neural network (CNN) based automated segmentation pipeline.
• Application of ensemble learning strategies, specifically Bagging and Augmenting, to enhance model performance.
• Quantitative evaluation of segmentation accuracy using metrics such as the Dice Similarity Coefficient (DSC).

Main Results:

• The automated pipeline achieved highly accurate segmentation of COVID-19 infected areas.
• Ensemble learning techniques demonstrated a positive impact on segmentation performance.
• The Bagging ensemble method yielded the highest Dice Similarity Coefficient, indicating superior segmentation accuracy.

Conclusions:

• Deep learning-based automated segmentation is effective for analyzing COVID-19 CT scans.
• Ensemble learning, particularly Bagging, can significantly improve the accuracy of COVID-19 lesion segmentation.
• This automated approach holds potential for efficient and reliable COVID-19 diagnosis and monitoring.