Uncertainty quantification in computed tomography pulmonary angiography.

Area of Science:

• Medical Imaging
• Radiology
• Computational Biology

Background:

• Computed tomography (CT) is crucial for pulmonary embolism (PE) detection.
• CT images can present artifacts that mimic PEs, complicating diagnosis.
• Distinguishing artifacts from true PEs is a significant challenge for radiologists.

Purpose of the Study:

• To develop a scalable hypothesis testing method for CT imaging.
• To introduce a Bayesian framework for quantifying uncertainty in PE detection.
• To assess the method's performance in high-noise and low-data scenarios.

Main Methods:

• A novel Bayesian framework was developed to quantify uncertainty.
• The framework focuses on compact structures potentially indicative of PE.
• Hypothesis testing was employed as a scalable approach for CT analysis.

Main Results:

• The Bayesian framework successfully quantifies uncertainty for potential PEs.
• The method demonstrates robustness in high-noise environments.
• The approach is effective even with limited or insufficient data.

Conclusions:

• The proposed Bayesian method offers a quantifiable approach to PE uncertainty.
• This technique aids in differentiating CT artifacts from actual pulmonary embolisms.
• The scalable hypothesis testing framework shows promise for improving diagnostic accuracy in challenging imaging conditions.