hongwang01/acdnet

【IJCAI2022】Adaptive Convolutional Dictionary Network for CT Metal Artifact Reduction

acdnet's Issues

请问这个和OSC-Net是一样的代码吗？

关于字典D的疑惑

作者您好，首先非常感谢您提供了这么出色的论文和代码！，有一个地方我想要请教一下，一个就是关于固定字典D，可以详细说说字典D的具体来源和生成方法吗？

There is a question about metal

First of all, thanks for your contribution to the network research.

It works very well in most cases but we found some problem.
First,If a patient who Suffer from cerebral hematomas need to Wear Surgical head stand. there have some CT scans image before this patient's operation which shown two or three kind of metals that belonging to different materials. It is hard to apart them for creating mask.if we use the same way which like CNN-MAR to prepare train or test data. It will have some serious problems because that is totally differently about real image.

Could you give some proposal about what should we do for this?

Secondly, we found that when using natural images to make metal artifact images, some metals are not at the anatomical position they should be at and this situation could be found in lots of work and research.Even in CNN-MAR,they said that the mask should at the anatomical position they suit be at.

For training a model with a fixed relative position of metals, whether using masks will affect the network performance？

looking forward to your reply and wishing you success in your research.