In theory yes. The first issue is knowing the follower count of an npub. They are all just aggregated from various relays. Then you need to keep track of these somewhere. 

And then it really depends on the time window you are looking at. The currently popular notes DVM only looks at one hour, which I found to show a wide range of diverse accounts and why i went for it. On a day or longer scale (see primals DVMs, or nostr band) it is true that large accounts make it top trending all the time (naturally, larger audience, more interactions etc).  

But yes, once you collect reliable follower count and store it, you could divide the interactions by that. Then you probably bias in the other direction, because larger accounts also have more "dead followers". Ideally you pick active followers. But we kind of enter a high processing environment doing this for every npub 🙈

That being said, yes if resources are available, it's totally doable.