I have a set of CT and MRI images that I aim to use for a classification task (disease presence: yes/no). The CT and MRIs do not correspond to the same patient but for a different set of patients (some have CT and some MRIs). Can I use a multi-input network in this case? Do multi-input networks should have 1:1 correspondence?
$\begingroup$ What would be the intended decision of CT belongs a person with disease, but MRI does not? $\endgroup$– gunesMar 25 at 0:01
$\begingroup$ no, you can use any input whatsoever, check multimodal neural networks, are an excellent example of this $\endgroup$– Alberto SinigagliaMar 27 at 12:57