A proper answer to this question is too extense to be written here. Molecular imprinting technology can be performed from different approaches: covalent and non-covalent, organic or inorganic, etc. Each of them has its own advantages and flaws. To select one or other depends on several parameters, per example whether you need fast response or swelling of your sensing material can be a problem. It also depends on the kind of transducer that you are going to use (e.g. if you want to use optic techniques you probably need transparent or non luminescent materials). I would recommend to have a look on the literature. You can find several good reviews on the topic, from quite general ones to more specific works.
Please take a slightly crosslinked polymer, add your substrate and then continue cross linking to maximum degree. Finally, remove the substrate via extraction. Then you have the required imprinted matrix.
The polymers themselves can be used to selectively adsorb an intended ligand as synthetic polymeric receptor sites, then you need a method of detection (binding detection and signal processing and amplification). It seems people have been working with various different methods, though lately QCM seems to be an area of interest. I'll throw in some links to look at regarding MIP based sensing using QCM and piezoelectric microgravimetry chemosensors. As suggested, see what the literature presents, a good starting point might be in form of any of many excellent review regarding the topics of imprinting and imprinting for biosensing and working your way from there.