Singularity extraction is a common prerequisite in BEM to evaluate singular integrals by separating the singular part of the singular integral from the regular part. The separation is obtained by truncating a series expansion of the singular kernel at the singular point. The singular part is then computed analytically, while a suitable quadrature rule is applied on the regular part. Nearly singular integrals are regular integrals, but with a very strong oscillation due to nearby presence of the singularity (e.g., when the singular point is very near the integration domain). For line integrals, the singularity extraction was generalized to nearly singular kernels by writing the expansion about the nearby complex points. In this talk we present this idea in the isogeometric framework to model problems on non-smooth 2D domains. In numerical examples we demonstrate that the expected optimal orders of convergence for the approximate solutions of the given boundary value problems are achieved with a small number of uniformly spaced quadrature nodes, using a quasi-interpolation-based quadrature rule.