Patient Identification Based on Deep Metric Learning for Preventing Human Errors in Follow-up X-Ray Examinations.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biometrics

Background:

• Patient identity verification is crucial for accurate medical records.
• Current methods for identifying patients from clinical images face performance limitations due to image variability.
• Deep learning offers potential for improving the robustness of these identification systems.

Purpose of the Study:

• To propose a novel deep learning-based method for automatic patient identification using chest X-ray images.
• To enhance the accuracy and reliability of patient verification in clinical settings.
• To address the challenges posed by image variability in automated identification systems.

Main Methods:

• A deep metric learning approach utilizing a deep convolutional neural network (DCNN) with an EfficientNetV2-S backbone was developed.
• The method involved three stages: preprocessing, DCNN feature extraction, and deep metric learning classification.
• Training was performed on the NIH chest X-ray dataset (ChestX-ray8).

Main Results:

• The proposed method achieved a high area under the receiver operating characteristic curve (0.9894) and a low equal error rate (0.0269) on the PadChest dataset.
• A 1280-dimensional feature extractor pretrained for 300 epochs demonstrated optimal performance.
• Top-1 accuracy reached 0.839 on a dataset including both PA and AP chest X-ray views.

Conclusions:

• The developed deep learning method shows significant promise for automated patient identification from chest X-rays.
• This technology can help reduce medical malpractice stemming from human errors in patient misidentification.
• The findings support the integration of advanced AI techniques into Picture Archiving and Communication Systems (PACS).