|
The IFF-NeuroSlides Consortium brings together the University of Michigan’s Machine Learning in Neurosurgery Lab and Ian’s Friends Foundation to assemble the most comprehensive digital pathology resource ever created for pediatric brain tumors.
Leveraging Ian’s Friends Foundation’s tissue-banking network, we are digitizing and harmonizing >20,000 whole-slide images (WSIs) from more than 2,000 children across the full WHO diagnostic spectrum — from diffuse and circumscribed gliomas to embryonal, pineal, and metastatic tumors.
Each case is accompanied by a rich panel of histochemical and immunohistochemical stains, and our pipeline includes algorithmic stain alignment that lets us analyze multi-stain biology as a single 3-D “stack” of data points. These slides, together with corresponding MRI and curated clinical metadata, form the backbone of a secure, de-identified research dataset ready for large-scale AI development.
Building on this unprecedented dataset, we are training a suite of state-of-the-art deep-learning models under the umbrella name NeuroSlides-AI.
Core objectives include:
(i) coarse-to-fine tumor subtype classification,
(ii) cross-stain representation learning that transfers knowledge from rare IHC stains back to routine H&E,
(iii) prediction of driver mutations and patient survival directly from tissue morphology, and
(iv) “virtual staining” and automated pathology-report generation to streamline clinical workflows.
By unifying histology with MRI and outcome data, the consortium aims to accelerate biomarker discovery, refine treatment stratification, and ultimately improve the lives of children with brain tumors—precisely the mission at the heart of Ian’s Friends Foundation.
| 
