Efficient multimodal deep-learning-based COVID-19 diagnostic system for noisy and corrupted images.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• COVID-19 diagnosis relies on efficient technologies for mitigation and detection.
• Deep learning models (DLMs) show promise for COVID-19 detection but are sensitive to data corruption.
• Publicly available datasets for DLMs are often adulterated, impacting model performance.

Purpose of the Study:

• To enhance COVID-19 detection using multimodal diagnostic systems with adulterated chest X-ray images.
• To develop a deep learning model (DLM) robust to corrupted and noisy data.
• To improve the reliability of AI in medical diagnostics.

Main Methods:

• Proposed a deep learning model (DLM) with convolutional and pooling layers for feature extraction.
• Utilized a batch normalization layer to prevent overfitting in the DLM.
• Employed two multimodal diagnostic systems for COVID-19 detection from adulterated chest X-ray images.

Main Results:

• Achieved an average accuracy of 98% in detecting normal, COVID-19, and viral pneumonia cases.
• Demonstrated high performance even with corrupted and noisy chest X-ray images.
• Matched the performance of standard techniques reported in existing literature.

Conclusions:

• The developed diagnostic systems exhibit robustness in detecting COVID-19.
• This robustness is vital for real-world applications with imperfect data.
• The study highlights the potential of AI in overcoming data quality challenges in medical diagnostics.