MFDNN: multi-channel feature deep neural network algorithm to identify COVID19 chest X-ray images.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Chest X-ray imaging (CXI) is crucial for diagnosing COVID-19.
• Accurate and efficient detection of SARS-CoV-2 is vital for patient care and public health.
• Existing deep learning models face challenges with unbalanced datasets and feature extraction.

Purpose of the Study:

• To develop and evaluate a Multi-channel Feature Deep Neural Network (MFDNN) algorithm for COVID-19 detection using CXI.
• To improve the accuracy and efficiency of COVID-19 screening through advanced deep learning techniques.
• To address data imbalance issues in medical datasets.

Main Methods:

• Implementation of a Multi-channel Feature Deep Neural Network (MFDNN) algorithm.
• Integration of data over-sampling technology to handle unbalanced datasets.
• Utilizing multi-channel feature fusion for enhanced feature extraction and diagnostic accuracy.

Main Results:

• The MFDNN model achieved an average test accuracy of 93.19% across all datasets.
• MFDNN demonstrated superior precision, recall, and F1 Score compared to traditional deep learning models (VGG19, GoogLeNet, Resnet50, Desnet201).
• Ablation experiments confirmed the superiority of multi-channel CNNs and the necessity of each MFDNN component.

Conclusions:

• The proposed MFDNN algorithm offers a highly accurate and efficient method for COVID-19 detection from chest X-rays.
• MFDNN effectively addresses data imbalance and enhances feature extraction, leading to improved diagnostic performance.
• The study validates the effectiveness of multi-channel approaches in deep learning for medical image analysis.