MCSC-Net: COVID-19 detection using deep-Q-neural network classification with RFNN-based hybrid whale optimization.

Area of Science:

• Medical Imaging and Artificial Intelligence
• Deep Learning in Healthcare
• Radiology and Diagnostic Imaging

Background:

• Accurate COVID-19 diagnosis is critical for patient outcomes and public health but is often time-consuming and requires expert interpretation.
• Existing deep learning (DL) models struggle with the accurate diagnosis of COVID-19 and other respiratory conditions using chest X-rays (CXRs).
• There is a need for advanced DL models capable of analyzing low-radiation imaging like CXRs for efficient and precise disease detection.

Purpose of the Study:

• To develop and evaluate a novel multi-class CXR segmentation and classification network (MCSC-Net) for the accurate detection of COVID-19.
• To improve upon the diagnostic accuracy limitations of current DL models in identifying COVID-19 and other lung pathologies from CXR images.

Main Methods:

• The MCSC-Net employs a hybrid median bilateral filter (HMBF) for noise reduction and enhancement of infected regions in CXRs.
• Segmentation of COVID-19 regions is performed using a skip connection-based residual network-50 (SC-ResNet50), followed by feature extraction with a robust feature neural network (RFNN).
• A disease-specific feature separate attention mechanism (DSFSAM) and the Hybrid whale optimization algorithm (HWOA) are utilized for distinct feature extraction and selection, with classification performed by a deep-Q-neural network (DQNN).

Main Results:

• The MCSC-Net achieved high classification accuracies: 99.09% for 2-class, 99.16% for 3-class, and 99.25% for 4-class CXR image analysis.
• These results demonstrate superior performance compared to existing state-of-the-art approaches in multi-class lung disease classification from CXRs.

Conclusions:

• The MCSC-Net effectively performs multi-class segmentation and classification of CXR images with high accuracy.
• This AI-driven approach shows significant promise for integration into future clinical practice, complementing existing diagnostic methods for patient evaluation.