BioinformaticsCancer SciencePeer-reviewed
Label Noise in Pathological Segmentation is Overlooked, Leading to Potential Overestimation of Artificial Intelligence Models
K. Harada, Yuichiro Nomura, Daisuke Komura…·Shingo Sakashita
This study investigates how label noise in pathology segmentation affects AI model performance, finding that boundary-altering noise leads to overestimation of model accuracy.