在卷积神经网络的训练过程中,将标签不确定性纳入训练过程中,可以提高多巴胺载体SPECT中某些和不确定的病例之间的歧视性能

Aleksej Kucerenko1, Thomas Buddenkotte2, Ivayla Apostolova2

  • 1xAILab Bamberg, Chair of Explainable Machine Learning, Faculty of Information Systems and Applied Computer Sciences, Otto-Friedrich-University, Bamberg, Germany.

概括

将读者不确定性纳入深层卷积神经网络 (CNN) 对多巴胺载体 (DAT) -SPECT解释的训练,可以改善对某些不确定的病例的歧视,而不会影响整体准确性.