An oppositional-Cauchy based GSK evolutionary algorithm with a novel deep ensemble reinforcement learning strategy...

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Computational Biology

Background:

• The COVID-19 pandemic necessitates rapid and accurate diagnostic tools.
• X-ray imaging is a crucial modality for accelerated COVID-19 diagnosis.
• Existing diagnostic methods require enhancement for efficiency and accuracy.

Purpose of the Study:

• To develop an automated artificial intelligence (AI)-based framework for analyzing X-ray images to detect COVID-19.
• To optimize deep convolutional neural networks (CNNs) for improved COVID-19 classification accuracy.
• To enhance diagnostic efficiency by reducing the number of required classifiers through a selective ensemble approach.

Main Methods:

• Designed a novel AI framework integrating an ensemble of deep CNNs.
• Optimized CNN architectures using a modified gaining-sharing knowledge (GSK) algorithm with opposition-based learning (OBL) and Cauchy mutation.
• Employed a deep Q network for selective ensemble classification, combining reinforcement learning (RL) with optimized CNNs.

Main Results:

• Achieved high diagnostic performance on two public COVID-19 X-ray datasets.
• Demonstrated significant accuracy (0.9914) and AUC (0.9903) on the Kaggle dataset.
• Showcased excellent accuracy (0.9877) and AUC (0.9884) on the Mendely dataset, validating the model's effectiveness.

Conclusions:

• The proposed deep RL-optimized ensemble approach effectively distinguishes COVID-19 patients from healthy individuals using X-ray images.
• The AI framework offers a promising solution for accelerated and accurate COVID-19 diagnosis.
• This automated system can significantly support healthcare professionals in managing the pandemic.