Abstract : Following intravenous contrast injection, Dynamic Contrast Enhanced Computed Tomography (DCE-CT) allows access to tissue perfusion parameters. Unfortunately, safety concerns limit strongly the X-ray in DCE-CT, which produces noisy images hardly usable for direct evaluation of tissue enhancement with a spatial resolution that preserves spatial heterogeneity within tumors. Based on statistical multiple hypothesis testing, a new denoising algorithm for DCE-imaging sequences is proposed. Its main interest consists in preserving the enhancement structures typical of microvascular behaviors, important for diagnosis. This is achieved by mixing a spatial local approach for aggregation of voxels and a time-global statistical test procedure to separate the tissue dynamics. Applied to DCE-CT sequences, this new algorithm shows its capacity not only to preserve organ shapes but also to distinguish and denoise tissue enhancements even for small vessels or tumor structures. In a second step, using the denoised sequence, the same tests are used to build unsupervised and automatic tissue clustering. This clustering allows to differentiate, up to pixel level, tissues without any prior knowledge on their number.