I commend Jong and colleagues1 for addressing the scarcity of domain-specific resources in endoscopic artificial intelligence (AI) and for making GastroNet-5M, which is a rare, large-scale multicenter dataset, publicly available. Their comprehensive benchmarking across diverse downstream tasks underscores its strong potential to advance the field. Although the findings are enlightening, I believe several issues should be addressed.
I commend Jong and colleagues1 for addressing the scarcity of domain-specific resources in endoscopic artificial intelligence (AI) and for making GastroNet-5M, which is a rare, large-scale multicenter dataset, publicly available. Their comprehensive benchmarking across diverse downstream tasks underscores its strong potential to advance the field. Although the findings are enlightening, I believe several issues should be addressed.