Is there a way to combine Test-Time Augmentation (TTA) with sliding window inference?

[Guoxt]

Is there a way to combine Test-Time Augmentation (TTA) with sliding window inference? Specifically, how can TTA be implemented during sliding window inference, and how can the predictions be aggregated using methods such as averaging or others?

[Guoxt]

Means I want to implement sliding window segmentation prediction in TTA.