TCGA-LUNG数据预处理过程的相关问题

[zmcwsd]

非常感谢你们的工作，我想请教一下，1.TCGA-LUNG数据预处理过程，提取的patch size是多大，在那个倍率下提取的。 2.提供的LUNG的feature vector 是基于ResNet18的还是ResNet50，3.特征提取器是否经过自监督对比训练。

[K1seki221]

Thanks@zmcwsd for your attention. Please refer to the configuration of data preprocessing in DSMIL, we simply implemented the same setting.

[zmcwsd]

Thanks，但是我还是有些疑问，再次感谢你们的工作，DSMIL使用的是ResNet18，并且使用自监督训练的方式，作为特征提取器， 但是你们的论文里面的结果写的是ResNet50，并且在数据处理部分说使用的DSMIL，这有一点矛盾

感谢您的解答！

[K1seki221]

Thanks @zmcwsd for your dedication. It seems we didn't correctly title [Table 1] in our paper. Table 1 reports the results of ResNet-50 on Camelyon16, and ResNet-18 as the same as DSMIL on TCGA-NSCLC. We will soon revise it. Thanks again!