Lens Identification to Prevent Radiation-Induced Cataracts Using Convolutional Neural Networks

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Ionizing radiation exposure during computed tomography (CT) poses a risk of cataract formation.
• Technologist positioning and maneuvers are critical to minimize lens radiation but can be challenging to maintain consistently.
• Manual feedback methods for quality improvement are often cumbersome and less effective.

Purpose of the Study:

• To develop and evaluate deep learning models for automatic lens detection in CT examinations.
• To facilitate continuous, automated feedback for technologists to improve radiation safety practices.
• To enhance patient care by reducing radiation exposure to the lenses.

Main Methods:

• Utilized convolutional neural networks (CNNs) for developing automated lens detection models.
• Integrated models to provide real-time, continuous feedback to CT technologists.
• Focused on improving technologist performance in minimizing lens radiation exposure.

Main Results:

• Demonstrated high-performance characteristics of deep learning models in detecting lenses.
• Enabled automatic and continuous feedback, aiding technologist compliance with radiation safety protocols.
• Identified opportunities for addressing operational and process-based challenges in CT imaging.

Conclusions:

• Deep learning offers a viable solution for automatic lens detection and continuous quality improvement in CT.
• Automated feedback systems can enhance technologist performance and patient safety.
• These models have potential applications in population health research and other imaging tasks.