Vgg16 as the perceptual loss function on 1 channel data

Hi experts, I have one channel data tensor of raw MRI slice i.e. [16,1,320,320] and I want to use Vgg16 as the perceptual loss function. So If I am not wrong, we push the output and the ground truth inside the vgg and compare their activations after relu using a loss function like MSE, so basically comparison the loss between the activation maps right. So vgg, as a pretrained network, works on the 3 channel images but my input tensor is 1 channel. Should I just copy the 1 channel into 3 channels? I need some guidance in this. Thanks!