Multi-Energy Evaluation of Image Quality in Spectral CT Pulmonary Angiography Using Different Strength Deep Learning...

Area of Science:

• Medical Imaging
• Radiology
• Computed Tomography

Background:

• Dual-energy CT pulmonary angiography (DECT-PA) is crucial for diagnosing pulmonary embolism (PE).
• Image quality in DECT-PA impacts diagnostic accuracy, necessitating advanced reconstruction techniques.
• Deep learning spectral reconstruction (DLSR) offers potential improvements over standard methods.

Purpose of the Study:

• To compare image quality of virtual monochromatic images (VMIs) generated by standard versus strong DLSR.
• To evaluate different energy levels (35-200 keV) of VMIs reconstructed using DLSR on DECT-PA.

Main Methods:

• Retrospective analysis of 70 DECT-PA scans (15 with PE, 55 without).
• Reconstruction of VMIs at various energy levels using standard and strong DLSR.
• Quantitative assessment of attenuation, SNR, and CNR; qualitative assessment at 40-100 keV.

Main Results:

• Strong DLSR demonstrated significantly lower image noise and higher cardiovascular SNR and CNR compared to standard DLSR (p < 0.0001).
• PE-specific CNR was significantly higher with strong DLSR (p < 0.0001).
• Subjective image quality was diagnostically acceptable across evaluated keV levels with both DLSR types.

Conclusions:

• Strong DLSR significantly enhances image quality in DECT-PA.
• Improved SNR and CNR with strong DLSR contribute to better diagnostic performance.
• Strong DLSR represents an advancement for DECT-PA image reconstruction.