Diffusion weighted imaging is prone to artefacts. Sources including hardware instability, bulk motion and cardiac pulsation can all induce spurious signal intensities (i.e. corruption) which negatively affect derived measurements. To combat this an additional processing step may be added to detect (and subsequently reject) such corrupted data points; however, with the increasing use of multi-shell acquisitions a number of existing approaches (constrained to a single b-value shell and/or unsuitable diffusion tensor models) are no longer applicable, limiting available choices. With this abstract we propose a new multi-shell detection algorithm and provide preliminary experimental results.