I have some dicom files about hearts. It has 50 time frames and 10 slices for a individual's heart.
So the data matrix shape is 208(row number)*186(columns number)*6(slice number)*50(time frame number).
Here I want to transform this matrix to a nifti format. I can transform every slice to a nifti format with using dcm2nii.
But how to transform these 10 slices to a nifti file all together. My procedure need a 208*186*6*50 matrix to label cavities. Which parameter should I use to do complete this object?

what happens if you run the command 'dcm2niix -f %s_%p_%t /path/to/DICOM'? All dcm2niix does is stack series as defined by the DICOM images. If your data is all part of the same series, all the slices should be combined into a single file. If your data is acquired as different series, you could merge your NIfTI images with fslmerge. The output provided by dcm2niix typically provides insights into the properties that caused images to be stacked or separated (e.g. for MRI data echo time, coil used, slice orientation, etc. can all be reasons slices from the same series are not merged). Sometimes using the parameter "-m y" or "-m n" may change how slices are combined, but this is usually a Hail Mary for poorly specified DICOM data.

Sorry for the delay....
It looks like my dataset are comprise  of different series. This is file structure showed by MicroDicom viewer:

All patients(Ptients:1, Images: 1453)
XXXXXXXX
YYYY heart(MR: 15 series)
    CINE segmented_SAX_b1
    CINE segmented_SAX_b2
    CINE segmented_SAX_b3
    CINE segmented_SAX_b4
    CINE segmented_SAX_b5
    CINE segmented_SAX_b6
    CINE segmented_SAX_b7
    CINE segmented_SAX_b8
    CINE segmented_SAX_b9
    CINE segmented_SAX_b10
    CINE segmented_SAX_b11
    CINE segmented_SAX_b12
    CINE segmented_SAX_b13
    CINE segmented_SAX_InlineVF
    Inline_VF_Results

This is a heart short axis MRI data.
The different series mean different location heart slices with each slice having 50 time frames.
should I merge my NIfTI images transformed by dcm2nii with fslmerge?

Your data are explicitly stored in different series. Therefore, the DICOM images are explicitly instructing dcm2niix to store these as separate images. You could use fslmerge or a simple Matlab script to combine series.

I have searched the fslmerge and found a R function https://rdrr.io/cran/fslr/man/fslmerge.html and a python function https://pythonhosted.org/nipype/interfaces/generated/nipype.interfaces.fsl.utils.html . Are these function equal?
It looks like all functions do are concatenating these matrix to a new matrix with a additional dimension. Is this right?

Yes.