Yes, it's possible, but it's likely going to be far more work than you're thinking and in the end you'll likely be left with very little of your engine. So it's probably a bad idea.

Unreal is a tightly-integrated package, it's not designed to be used piecemeal, and it's very hard to decouple its various major subsystems from eachother. The component you want to use, the rendering engine, is going to depend on or talk with the core object model and garbage collection, the asset serialization system, the physics layer, the editor for building materials (and in turn the rest of the editor infrastructure), et cetera.

You will almost certainly be better off ripping out the parts of your engine you really want to use, the parts that Unreal is not currently providing analogous subsystems for, and extending UE as you noted (with plugins, et cetera) to leverage that code you've already written for gameplay, et cetera. You will likely need to adapt that code somewhat to work in the UE model; for example if you have a mechanism for representing in-game objects or entities you'll likely need to discard that in favor of using Unreal's AActor classes, et cetera.

Yes, it's possible, but it's likely going to be far more work than you're thinking and in the end you'll likely be left with very little of your engine. So it's probably not the best idea.

Unreal is a tightly-integrated package, it's not designed to be used piecemeal, and it's very hard to decouple its various major subsystems from eachother. The component you want to use, the rendering engine, is going to depend on or talk with the core object model and garbage collection, the asset serialization system, the physics layer, the editor for building materials (and in turn the rest of the editor infrastructure), et cetera.

You'd be better off taking your existing gameplay code you want to keep, and adapting that to Unreal (not everything in Unreal has to be done with Blueprints, you can write native C++ just fine, directly or via plugins, as you've noted). You will need to adapt that gameplay logic to work within the framework of Unreal (e.g., AActor instead of your own existing game object/entity system, and so on), so you'll be abandoning almost all of your "engine" code except perhaps systems that Unreal provides no analogue for.

This will still be a fairly significant effort, potentially (depending on the size of your gameplay code base). It's up to you to decide if the perceived gain in graphics functionality is worth the cost.

Remember that good graphics involves good art, as well, and art that is tailored to the way the graphics code works. It's possible your existing art will need to be re-authored because it's flat-out incompatible or because it just doesn't look good within Unreal's lighting/rendering models, et cetera. Keep in mind that overheard as well.

I am very skeptical this will be a net win for you.