Regression Using Logs of Variables

I have converted 80 ("Count") pairs of X and Y to L10(X) and L10(Y) and performed a regression analysis. Assuming a valid correlation, I know I can predict L10(Y) from a given L10(X) to equal the Y intercept plus slope multiplied by L10(X). I can then convert the predicted L10(Y) to a predicted actual Y by using 10^L10(Y).

My question has to do with determining what I call my confidence interval (may be wrong term). For a confidence interval of 98%, I use the following Excel function: TINV( (1-98%), (Count - 2)) * SQRT( 1/Count + (given L10(X) - average L10(X)) ^2 / SSX). Is the confidence interval produced by this equation an L10()? If yes, then must I convert it to an actual number to compare it to my predicted actual Y?