Shouldn't use resnet for not image type data?

i have a data that it is ROICorrelation matrix from fmri image. i fed it into resnet and got a good performance but I don’t know if it is right to use resnet for nonimage data or not?

While the original use case was an ImageNet classification, I wouldn’t see any valid points against training or fine-tuning this model to your use case.
It also seems to work, which is great. What is your concern about using this model?

HI. thank you for your response.
my concern is that my input data isn’t image is kind of extracted feature from fmri images.