BriefGPT.xyz
Sep, 2024
对全切片图像分类中多实例学习可靠性的定量评估
Quantitative Evaluation of MILs' Reliability For WSIs Classification
HTML
PDF
Hassan Keshvarikhojasteh
TL;DR
本研究解决了现有多实例学习(MIL)模型在全切片图像(WSIs)分类中缺乏可靠性评估的问题。论文提出了三种度量方法来比较MIL模型的可靠性,并通过区域标注数据集进行测试。研究发现,尽管均值池化实例(MEAN-POOL-INS)模型在设计上较为简单,但其可靠性优于其他网络,具有重要的应用潜力。
Abstract
Reliable models are dependable and provide predictions acceptable given basic domain knowledge. Therefore, it is critical to develop and deploy reliable models, especially for
Healthcare Applications
. However,
Multiple
→