Pulmonary Embolism Detection in Unenhanced CT Exams: A Radiomic-Based Classifier as a Promising Screening-Diagnostic...

Area of Science:

• Medical Imaging
• Radiomics
• Pulmonary Embolism Diagnostics

Background:

• Pulmonary embolism (PE) is a critical condition requiring timely diagnosis and treatment.
• Current radiomics tools for PE detection are limited to CT pulmonary angiography (CTPA).
• The utility of radiomics for PE detection on non-contrast CT remains uninvestigated.

Purpose of the Study:

• To explore the potential of radiomics for identifying pulmonary emboli on unenhanced chest CT scans.
• To develop and validate a radiomics-based classifier for thrombus detection.
• To assess the diagnostic performance of radiomics in detecting PE without contrast.

Main Methods:

• Retrospective analysis of 57 CTPA examinations.
• Extraction of 836 radiomic features from embolus and normal-flow pulmonary artery regions of interest (ROIs).
• Development of a logistic regression classifier using statistically significant radiomic features.

Main Results:

• 242 wavelet-filtered radiomic features showed significant differences between embolus and normal ROIs.
• The classifier achieved an AUC of 0.87 and an accuracy of 0.77 in the validation cohort.
• The model demonstrated high diagnostic performance with minimal misclassifications.

Conclusions:

• Radiomics shows potential for identifying thrombi on unenhanced chest CT.
• This approach offers a promising method for early acute PE detection.
• Further validation in larger cohorts is warranted to support clinical application.