The problem of determining invariance kernels for planar single-input nonlinear systems is considered. If K is a closed set, its invariance kernel is the largest subset of K with the property of being positively invariant for arbitrary measurable input signals. It is shown that the boundary of the invariance kernel is a concatenation of solutions of two so-called extremal vector fields. Moreover, only the solutions through a finite number of special points are of interest. This result makes it possible to devise an algorithm which determines the invariance kernel of a simply connected set in a finite number of steps.