Enhancing diagnostic safety with low iodine, low radiation CTPA classification using deep learning

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Pulmonary embolism (PE) diagnosis relies on CTPA, which uses high iodine contrast and radiation doses.
• High contrast and radiation doses raise concerns for renal injury and patient safety.
• Developing safer CTPA protocols is crucial for clinical practice.

Purpose of the Study:

• To develop and evaluate a deep learning framework for accurate PE diagnosis using low-iodine and low-radiation CTPA.
• To improve diagnostic performance in PE detection under reduced contrast and radiation conditions.
• To provide a practical solution for enhancing safety in clinical CTPA imaging.

Main Methods:

• A two-stage deep learning framework was proposed, integrating image enhancement and classification.
• The framework leveraged original low-exposure images and their super-resolved counterparts.
• A public low-iodine, low-radiation CTPA dataset was created to support research.

Main Results:

• The proposed deep learning framework achieved an AUC of 0.928.
• The method demonstrated substantial improvement over single-branch baselines.
• Balanced sensitivity and specificity were maintained, indicating robust diagnostic performance.

Conclusions:

• The deep learning framework enables accurate and safer PE diagnosis with reduced contrast and radiation.
• This approach offers a practical solution for improving patient safety in CTPA.
• The publicly released dataset facilitates further research in safe medical imaging.