Enhancing InceptionResNet to Diagnose COVID-19 from Medical Images.

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Deep Learning

Background:

• Current COVID-19 diagnosis models using X-rays often overlook crucial performance metrics.
• Key parameters like precision, sensitivity, specificity, F1-score, and ROC-AUC are vital for accurate model assessment.
• The proposed Enhanced InceptionResNet aims to address these limitations by incorporating advanced deep learning techniques.

Purpose of the Study:

• To develop and evaluate an improved deep learning model for COVID-19 diagnosis using chest X-ray images.
• To compare the performance of the Enhanced InceptionResNet against traditional ResNet and InceptionResNet models.
• To assess the model's effectiveness using comprehensive performance metrics beyond simple accuracy.

Main Methods:

• Three deep learning models were utilized: ResNet, InceptionResNet, and the novel Enhanced InceptionResNet.
• The models were trained and validated on a balanced dataset of 2600 chest X-ray images.
• Performance evaluation included accuracy, loss, confusion matrix analysis, precision, recall, F1-score, and ROC-AUC.

Main Results:

• The Enhanced InceptionResNet significantly outperformed both ResNet and InceptionResNet across all evaluated metrics.
• Achieved high validation and testing accuracy (99.0% and 98.35% respectively), demonstrating robust performance.
• Showcased superior feature extraction capabilities, leading to more reliable COVID-19 identification.

Conclusions:

• The Enhanced InceptionResNet is highly effective for COVID-19 diagnosis from chest X-rays, outperforming existing models.
• The model shows significant promise for broader medical image classification tasks.
• Future research should focus on larger datasets and hyperparameter optimization for further performance enhancement.