Question about multi-modality medical image.

[RitchieAlpha]

Dear author: Thank you for your solid research, we have benefited from the project a lot. We know that in many cases, the medical image dataset always has many modalities (or channels) like brats2018, etc. My question is how to convert the single modality data pre-train weight to a multi-modality pre-train? Is that a good way to replace the input channel of the first convolution layer, and fine-tune the network? Thanks, looking forward to your reply.

[zfy514]

尊敬的作者 ：感谢您的扎实研究，我们从该项目中受益匪浅。 我们知道，在许多情况下，医疗图像数据集总是有许多模式（或通道），如 brats2018 等。我的问题是如何将单一模式数据预训练权重转换为多模式预训练？这是更换第一卷积层的输入通道并微调网络的好办法吗？ 谢谢， 期待您的回复。 Hi ，Did you solve this problem?I also do related research. Thanks for your reply!