A Multiphasic CT-Based Decision Support Model for Pre-Treatment Differentiation of Pediatric Hepatoblastoma and Focal Nodular Hyperplasia

1

This study developed a multiphasic CT-based model to differentiate pediatric hepatoblastoma from focal nodular hyperplasia using deep learning and quantitative features.
2

The analysis included 612 children, with 523 diagnosed with hepatoblastoma and 89 with focal nodular hyperplasia, over a study period from 2011 to 2025.
3

The DL+QEF model achieved an AUC of 0.980 and 94.2% accuracy in the temporal test cohort, outperforming other models in sensitivity and specificity.
4

Junior radiologists improved their diagnostic accuracy from 88.4% to 95.0% when assisted by the DL+QEF model, particularly in atypical cases.
5

The study highlights the potential of the DL+QEF model to enhance CT interpretation and aid in the differentiation of pediatric liver lesions.

1