Deep learning denoising improves low-dose abdominal CT quality but not appendicitis detection accuracy

In low-dose abdominal CT (median 1.34 mSv), deep learning-based denoising significantly improved quantitative noise and overall image quality, with denoised filtered back projection (FBP) rated best subjectively, though sensitivity and specificity for acute appendicitis did not…

• Single-center retrospective study of 100 patients (median age 31) with suspected appendicitis scanned at median 1.34 mSv effective dose.
• Denoised IR achieved lowest quantitative noise; denoised FBP scored highest for overall subjective image quality.
• No significant difference in sensitivity or specificity for appendicitis diagnosis across FBP, IR, denoised FBP, and denoised IR series.

RadPigeon summaries are original and for information only. They are not clinical advice.