Combined clinical-radiomic-deep learning model accurately classifies pediatric pineal tumors on MRI

A multimodal model integrating clinical, radiomic, and deep learning features from MRI achieved a macro-average AUC of 0.912 for classifying pediatric pineal region tumors (germinoma, NGGCT, PPT), outperforming single-modality models.

• The study included 280 children: 94 germinomas, 119 non-germinomatous germ cell tumors (NGGCTs), and 67 pineal parenchymal tumors (PPTs).
• The model used radiomic and 2.5D deep learning features from T1WI, T2WI, and contrast-enhanced T1WI sequences alongside clinical variables.
• Model-assisted interpretation improved diagnostic accuracy for both junior and senior radiologists in a reader study.

Automated summary

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