NDL-Net: A Hybrid Deep Learning Framework for Diagnosing Neonatal Respiratory Distress Syndrome From Chest X-Rays

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neonatal Medicine

Background:

• Neonatal Respiratory Distress Syndrome (NRDS) is a critical condition in newborns.
• Accurate and timely diagnosis of NRDS is essential for effective treatment.
• Current diagnostic methods may have limitations in speed and accuracy.

Purpose of the Study:

• To introduce NDL-Net, a novel hybrid deep learning framework for NRDS diagnosis.
• To evaluate the diagnostic performance of NDL-Net using neonatal chest X-rays (CXR).
• To improve the early detection and management of NRDS in neonates.

Main Methods:

• Developed NDL-Net, a hybrid deep learning architecture.
• Integrated MobileNetV3 Large for image processing and ResNet50 for pattern detection.
• Incorporated a Long Short-Term Memory (LSTM) layer for temporal data analysis.

Main Results:

• NDL-Net achieved high diagnostic performance on neonatal CXR datasets.
• Achieved 98.09% accuracy, 97.45% precision, 98.73% sensitivity, 98.08% F1-score, and 98.73% specificity.
• Demonstrated low false negative and false positive rates, indicating superior diagnostic capability.

Conclusions:

• NDL-Net signifies a substantial advancement in medical diagnostics for NRDS.
• The framework enhances neonatal care through improved early detection of NRDS.
• NDL-Net offers a promising tool for the timely management of NRDS in newborns.