Posterior Fossa Tumor Detection​
Description
This multi-institutional study developed a deep learning model using MR imaging to detect and classify posterior fossa tumors in pediatric patients. The model, based on a modified ResNeXt-50 architecture, achieved high accuracy in tumor detection and classification, with performance comparable to radiologists. It demonstrated particularly strong results in predicting diffuse midline glioma of the pons and pilocytic astrocytoma, suggesting its potential to enhance radiologic diagnosis accuracy.
Creator
Seattle Children’s Hospital, Stanford University School of Medicine, Riley Children’s Hospital, Boston Children’s Hospital, Dayton Children’s Hospital, Lucile Packard Children’s Hospital, The Hospital for Sick Children, University of Utah School of Medicine
Information
Pediatrics or Adult
Pediatrics
Speciality
Oncology
Modality
MRI
Training
617 children (median age, 92 months; 56% males) from 5 pediatric institutions with posterior fossa tumors
Github
Publication
FDA