BraTS2024-preprocessing-labels-dataset.json

[TSbme]

I found that the default nnUNet preprocessing can only handle labels numbered 3.
In dataset.json,

"labels": {
    "background": 0,
    "whole tumor": [
        1,
        2,
        3
    ],
    "tumor core": [
        2,
        3
    ],
    "enhancing tumor": [
        3
    ]
},
"regions_class_order": [
    1,
    2,
    3
]

While the BraTS2024 dataset has labels [0, 1, 2, 3, 4].
How to handle the labels?
I wanna achieve the highest accuracy for ET segmentation.