Deep Learning-Based Decision-Tree Classifier for COVID-19 Diagnosis From Chest X-ray Imaging.

Area of Science:

• Artificial Intelligence in Medical Imaging
• Radiology and Diagnostic Imaging
• Infectious Disease Diagnostics

Background:

• The COVID-19 pandemic increased demand for rapid diagnostic tools.
• Reverse transcription polymerase chain reaction (RT-PCR) is definitive but time-consuming.
• Chest X-ray radiography (CXR) offers a faster, accessible alternative for identifying pneumonia.

Purpose of the Study:

• To investigate the feasibility of a deep learning-based decision-tree classifier for COVID-19 detection using CXR images.
• To develop an AI model for efficient pre-screening and triage of patients with suspected COVID-19.

Main Methods:

• A deep learning classifier using three binary decision trees was developed on the PyTorch framework.
• Convolutional neural networks (CNNs) formed the basis for each decision tree.
• The trees were trained to classify CXR images as normal/abnormal, identify tuberculosis, and detect COVID-19.

Main Results:

• The first decision tree achieved 98% accuracy in classifying normal versus abnormal CXR images.
• The second decision tree identified tuberculosis signs with 80% accuracy.
• The third decision tree demonstrated an average accuracy of 95% for detecting COVID-19.

Conclusions:

• The proposed deep learning-based decision-tree classifier shows high accuracy in detecting COVID-19 from CXR.
• This AI tool can assist in pre-screening patients, enabling faster triage and decision-making.
• Integrating AI with CXR can significantly enhance diagnostic workflows during pandemics.