If you assume he limps UTG with any two (doubtful, but lets work with it)

Then you are beat by:
33 88 AT QT KT T8 T3 AA

The last few hands are not real likely but not impossible that he slowplays AA this way.
Not sure why you rule out AT - this is not uncommon to limp UTG, and why would he bet the flop/raise the turn? Fear of flush draw, maybe - but you didn't bet it either.

There are 52 - 7 = 45 unknown cards, 45x44 = 1980 possible hands for villian

There are 6 x 3 = 18 PP hands (88/33/AA)

There are 4 x 2 = 8 higher sets (KT/QT)

There are 3 x 3 = 9 other boats (AT/T8/T3).

Under our assumptions, there is a 0.4% chance he has exactly a higher set, and overall a 1.7% chance you are beaten.

Now, our assumption is flawed fundamentally in that
a) He is not limping UTG with any 2
b) He is not raising you all in with any 2 most of the time.

So,cannot calculate the exact odds of his holding without more assumptions. Here are some hands that you beat than might make this play:

Nothing: Harrington's rule is 10% bluffs. Let'use that.
Other than the above hands that beat you:
AK/AQ/AJ/A8/A3
less likely
AX (X <= 9, not 2 pair)
KK
QQ

AK/AQ is probably pretty unlikely too, with the UTG limp. Actually, none of these seem real likely, but people can act silly, and for whatever reason, he might have thought your weak river bet was totally weak.

Let's maximize the chances you are good and say he makes this play with ANY ace (probably need a read here).

3 aces X 41 other cards (can't count the T) = 123
add in QQ/KK (played horribly) = 12 x 2 = 24

So we have: 147 hands you can beat
35 hands that beat you

So, you win 0.9 x 0.8 = .72 + .10 (bluff) 82% of the time, if he's a completly loose-agressive nut case.

Realistically - you are probably ahead more than 1/2 the time, so you can probably call.