边界加权 logit 一致性改善分割网络的校准
Boundary-weighted logit consistency improves calibration of segmentation networks
AI总结：
针对分割数据标签模糊导致的网络错误校准问题，提出边界加权的跨变换 logit 一致性正则化方法，在前列腺与心脏 MRI 分割上取得最先进校准效果。
摘要：
神经网络的预测概率与准确率通常仅呈弱相关。图像分割训练数据中固有的标签模糊性加剧了这种错误校准。我们表明，跨随机变换的 logit 一致性可充当空间变化的正则化项，防止在标签模糊的像素处产生过度自信的预测。我们对该正则化项的边界加权扩展在前列腺和心脏 MRI 分割中提供了最先进的校准性能。
英文摘要：
Neural network prediction probabilities and accuracy are often only weakly-correlated. Inherent label ambiguity in training data for image segmentation aggravates such miscalibration. We show that logit consistency across stochastic transformations acts as a spatially varying regularizer that prevents overconfident predictions at pixels with ambiguous labels. Our boundary-weighted extension of this regularizer provides state-of-the-art calibration for prostate and heart MRI segmentation.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
