SNELM: SqueezeNet-Guided ELM for COVID-19 Recognition

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• COVID-19 has caused millions of deaths globally, necessitating accurate diagnostic tools.
• Chest computed tomography (CT) offers precise imaging for COVID-19 diagnosis.
• Early and accurate diagnosis is crucial for patient management and disease control.

Purpose of the Study:

• To develop and evaluate a novel deep learning model for automated COVID-19 diagnosis from chest CT images.
• To assess the performance of the SqueezeNet-Extreme Learning Machine (SNELM) model in detecting COVID-19.
• To compare the proposed model's diagnostic accuracy against existing state-of-the-art methods.

Main Methods:

• Utilized two chest CT datasets (296 and 640 images).
• Applied extensive data augmentation techniques to enhance the training dataset.
• Developed a SqueezeNet (SN) model with complex bypass for feature extraction.
• Employed an Extreme Learning Machine (ELM) classifier with 2000 hidden neurons.
• Validated results using 10 runs of 10-fold cross-validation.

Main Results:

• The SNELM model achieved high diagnostic performance on both datasets.
• For the 296-image dataset: Sensitivity 96.35%, Specificity 96.08%, Precision 96.10%, Accuracy 96.22%.
• For the 640-image dataset: Sensitivity 96.00%, Specificity 96.28%, Precision 96.28%, Accuracy 96.14%.
• Performance metrics consistently exceeded those of seven other leading COVID-19 recognition models.

Conclusions:

• The proposed SNELM model demonstrates significant success in diagnosing COVID-19 from chest CT scans.
• The model's high accuracy and robustness suggest its potential as a valuable tool in clinical settings.
• SNELM offers a promising AI-driven solution for improving COVID-19 diagnostic capabilities.