AANS Online Scientific Session: Pediatrics-Machine Versus Human: Deep Learning to Automatically Detect and Diagnose Pediatric Brain Tumors in a Large Multi-Institutional Study

Machine Versus Human: Deep Learning to Automatically Detect and Diagnose Pediatric Brain Tumors in a Large Multi-Institutional Study

Video Transcription

Video Summary

In this video, Jennifer Kwan, a PGY5 resident at Stanford, discusses a deep learning model for diagnosing posterior fosso tumors. She explains that deep learning models are already being used in various fields, including medicine, and are effective in reading chest x-rays, EKGs, retinal scans, and stroke detection. The study focused on using artificial intelligence tools to diagnose posterior fosso tumors using imaging alone. The model architecture was based on a ResNxt50 pre-trained on ImageNet and fine-tuned using tumor images. The model's performance was tested on tumor and normal images it had never seen before. The model performed comparably to clinical experts and provided faster automated diagnosis. The team hopes to expand the model to other tumor types and incorporate genetic subtype information in the future.

Asset Subtitle

Jennifer Lauren Quon, MD

Keywords

deep learning model

diagnosing posterior fossa tumors

artificial intelligence tools

ResNxt50

automated diagnosis