In disease mapping, the aim is to estimate the spatial pattern in disease risk over an extended geographical region, so that areas with elevated risks can be identified. A Bayesian hierarchical approach is typically used to produce such maps, which represents the risk surface with a set of random effects that exhibit a single global level of spatial smoothness. However, in complex urban settings, the risk surface is likely to exhibit localized rather than global spatial structure, including areas where the risk varies smoothly over space, as well as boundaries separating populations that are geographically adjacent but have very different risk profiles. Therefore, this paper proposes an approach for capturing localized spatial structure, including the identification of such risk boundaries. The effectiveness of the approach is tested by simulation, before being applied to lung cancer incidence data in Greater Glasgow, UK, between 2001 and 2005.