There two parts to this:
Using the linear equation (model) to calculate a middle value for the response variable.
Calculating an upper and lower limit for this prediction using the spread of points (strength) of the scatterplot.