Real‑time COVID-19 diagnosis from X-Ray images using deep CNN and extreme learning machines stabilized by chimp...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Real-time COVID-19 detection from radiological images is crucial for rapid diagnosis.
• Traditional Deep Learning (DL) methods are time-consuming for training and parameter tuning.
• Extreme Learning Machines (ELMs) offer fast detection but require optimization for accuracy.

Purpose of the Study:

• To develop a novel, two-phase approach for efficient and accurate COVID-19 detection from chest X-ray images.
• To address the limitations of standard ELMs in image processing by optimizing their performance.
• To achieve real-time diagnostic capabilities for COVID-19.

Main Methods:

• A two-phase classification approach was designed, utilizing a deep Convolutional Neural Network (CNN) as a feature extractor.
• Extreme Learning Machines (ELMs) were employed for real-time detection in the second phase.
• The Chimp Optimization Algorithm (ChOA) was integrated to optimize ELM parameters, enhancing network reliability and performance.

Main Results:

• The proposed Chimp Optimization Algorithm-optimized ELM (ChOA-ELM) achieved high accuracy, reaching 98.25% on the COVID-Xray-5k dataset and 99.11% on the COVIDetectioNet dataset.
• Relative error was reduced by 1.75% and 1.01% compared to a standard convolutional CNN.
• Training time for the deep ChOA-ELM was exceptionally fast at 0.9474 milliseconds, with an overall testing time of 2.937 seconds for 3100 images.

Conclusions:

• The novel two-phase ChOA-ELM approach provides a highly accurate and real-time solution for COVID-19 detection using chest X-rays.
• This method significantly outperforms existing benchmarks, including classic DCNN, GA-ELM, CS-ELM, and WOA-ELM.
• The optimized ELM approach offers a promising direction for rapid and reliable medical image analysis in pandemic scenarios.