TensorBoardImageHandler for multiple branches

Is your feature request related to a problem? Please describe.
I have a problem using TensorBoardImageHandeler in 2 ways. I have have code snippet below. First, during each epoch, the index of the image is not fixed, so I always get new image and corresponding label and output.

Second, if I tried to add extra images, there is no way to point to with other labels. In case here, I want to change NC to NP branch.

        TensorBoardImageHandler(log_dir=log_dir, 
                                max_channels = 3,
                                batch_transform=from_engine(keys=["image", "label"]),
                                output_transform=from_engine_hovernet(keys=["pred"], nested_key=HoVerNetBranch.NC.value),
                               )

Describe the solution you'd like
Ideally , the image will be fixed to specific index, and I am able to point to different tag names for the handler, so extra branches can be incorporated.

Thanks A lot!

Start at the TensorBoardImageHandler entry point and trace how images, labels, outputs, and tags are selected across epochs. Define the expected behavior for a fixed image index and for separate tags or branches, then add coverage for both requested use cases before verifying the handler output.