Amit Gupta, Soroosh Mahboubi Global Radiology Scholar at Children’s Hospital of Philadelphia and Social Media Editor at European Radiology, shared a post on LinkedIn:
“Alex Towbin is one of the clearest voices in pediatric imaging AI.
His International Cancer Imaging Society presentation sparked the editorial I’m sharing today, and I’m truly honored to have co-authored it with him.
In our editorial, we explore why pediatric cancer imaging remains on the periphery of AI progress despite enormous potential.
The barriers are not technical alone:
- Rarity – pediatric cancers account for <1% of new diagnoses. Deep learning thrives on scale; pediatric oncology does not offer it.
- Fragmentation – even large centers may only accumulate a few hundred usable cases over decades.
- Restricted access – children represent <1% of publicly available medical imaging data.
- Heterogeneity – protocols, scanners, and contrast strategies vary widely across institutions.
And perhaps most importantly: pediatric AI is not small adult AI.
Children are not scaled-down adults. Their anatomy, tumor biology, and imaging appearances differ.
Adult-trained models applied to pediatric data can underperform, sometimes dangerously so.
And yet, there is reason for cautious optimism.
Early pediatric segmentation and detection studies show feasibility.
Radiomics and integrated imaging-omics approaches could meaningfully refine risk stratification and treatment monitoring if pursued with rigor and collaboration.
What this editorial ultimately argues is simple:
The question is no longer whether AI will enter pediatric cancer imaging.
It is whether we will build it transparently, collaboratively, and with the child — not the algorithm — at the center.
Alex Towbin, I’m grateful for your mentorship, generosity, and shared commitment behind this piece.
Some collaborations are academic.
Others are values-driven.
This one felt like both.”
Other Articles Featuring Amit Gupta on OncoDaily.