Optimizing pediatric chest CT: superior image quality and lower radiation dose with deep learning reconstruction

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Pediatric chest CT imaging requires balancing diagnostic image quality with radiation dose.
• Traditional hybrid iterative reconstruction (HIR) algorithms have limitations in dose reduction.
• Advancements in deep learning reconstruction (DLIR) offer potential for improved imaging protocols.

Purpose of the Study:

• To compare image quality and radiation dose between DLIR and HIR in unenhanced pediatric chest CT.
• To evaluate the effectiveness of DLIR in reducing radiation exposure for pediatric patients.
• To assess diagnostic image quality metrics using both subjective and objective assessments.

Main Methods:

• A hybrid prospective-retrospective study included 142 pediatric patients (<16 years).
• DLIR cohort used a new-generation scanner; HIR cohort was age- and indication-matched.
• Image quality assessed via radiologist Likert scales and quantitative noise/attenuation measurements; radiation dose via CTDIvol, DLP, and effective dose (ED).

Main Results:

• Subjective image quality was comparable between DLIR and HIR (κ=0.78).
• DLIR showed significantly lower image noise in lung parenchyma, trachea, and paraspinal muscle.
• Radiation exposure was markedly reduced with DLIR (e.g., median ED 0.70 vs 2.51 mSv, p<0.001).

Conclusions:

• DLIR enables substantial radiation dose reduction in pediatric chest CT.
• Diagnostic image quality is maintained or improved with DLIR compared to HIR.
• DLIR represents a promising technique for safer pediatric imaging.