Interesting set of charts in the LM2578A datasheet. I guess it's handy as quick reference, but seems a bit cumbersome beyond that, to be honest. It might be more useful to think in terms of %ΔI(L) instead of E-Top. At least, that's a more typical approach, in my experience.
You can use a wide range of inductance values and have the converter "work". If you use an inductor lower than the prescribed value per the datasheet, at a given set of load conditions, this just means that the inductor ripple current ΔI(L) will be higher. Conversely, if you use a higher inductance, than ΔI(L) will be lower. For a number of practical reasons, %ΔI(L) between 20 - 30% is usually selected as a reasonable compromise. In an ideal world, we'd want %ΔI(L) to be low, to keep voltage ripple low, switch peak currents low, and output capacitor ripple current low. But, if we selected a (high value) inductance based off of, say, %ΔI(L) = 5%, then we'd have a monstrously large inductor - usually larger than the given application can tolerate. So, we back down %ΔI(L) to somewhere in the 20 - 30% range (typically), to use a physically smaller inductor.
So, then you see that inductance value is proportional to physical size of the inductor, for a given application, if switching frequency and loading conditions are held constant. The reasons for this are probably beyond the scope of this discussion, but are related to saturation of the core material (most generally).