Ode So, Fujikawa Atsuko, Hiroishi Atsushi, Saito Yuki, Tanuma Takao, Suzuki Daigo, Sasaki Yuichi, Mimura Hidefumi
Department of Radiology, St. Marianna University School of Medicine, 2-16-1 Sugao, Miyamae-Ku, Kawasaki, Kanagawa, 216-8511, Japan.
Imaging Center, St. Marianna University School of Medicine, 2-16-1 Sugao, Miyamae-Ku, Kawasaki, Kanagawa, 216-8511, Japan.
Jpn J Radiol. 2025 Aug;43(8):1365-1371. doi: 10.1007/s11604-025-01775-9. Epub 2025 Apr 7.
To evaluate the impact of an AI-based, noise reduction technique for compensation of image degradation on pediatric and neonatal chest and abdomen radiography using a visual grading analysis.
Forty-six consecutive cases of pediatric and neonatal chest X-rays were identified for the quality evaluation. The images underwent AI-based noise reduction processing (Intelligent NR, Canon Inc.). All the images were randomized, and were evaluated from 1 to 4 for image quality by three board-certified radiologists in consensus. A score of "1" indicated the desired anatomy or features were not seen, "2" indicated quality between one and three, "3" indicated adequate quality, and "4" indicated higher than required image quality. A Wilcoxon signed rank test was used to assess the significant difference between images from conventional noise reduction versus those from the AI-based noise reduction.
The images processed with the INR(Intelligent NR) noise reduction had a higher image quality than the conventionally processed images, with a significant difference between the two groups (p < 0.05).
The AI-based noise reduction technique improved the image quality of pediatric and neonatal chest and abdominal radiography significantly.
使用视觉分级分析评估一种基于人工智能的降噪技术对小儿及新生儿胸部和腹部X线摄影图像退化的补偿效果。
选取46例连续的小儿及新生儿胸部X线病例进行质量评估。对图像进行基于人工智能的降噪处理（智能降噪，佳能公司）。所有图像随机排列，由三名获得委员会认证的放射科医生共同对图像质量从1至4进行评估。“1”分表示未见到所需的解剖结构或特征，“2”分表示质量在1至3之间，“3”分表示质量足够，“4”分表示图像质量高于要求。采用Wilcoxon符号秩检验评估传统降噪图像与基于人工智能的降噪图像之间的显著差异。
经智能降噪（INR）处理的图像比传统处理的图像具有更高的图像质量，两组之间存在显著差异（p < 0.05）。
基于人工智能的降噪技术显著提高了小儿及新生儿胸部和腹部X线摄影的图像质量。
