Biomedical imaging lacks label-free microscopy techniques able to reconstruct the contour of biological cells in solution , in 3D and with high resolution, as required for the fast diagnosis of numerous diseases. Inspired by computational optical coherence tomography techniques, we present a tomographic diffractive microscope in reflection geometry used as a synthetic confocal microscope, compatible with this goal and validated with the 3D reconstruction of a human effector T lymphocyte. Optical diffraction microscopy is an important tool in biological and biomedical imaging as it can be used on live cells and does not require staining. Yet, its poor axial resolution compared to the transverse one limits its interest for three-dimensional (3D) imaging. Now, an increasing number of applications would benefit from highly resolved 3D images of cells. In particular, the ability to observe the contour deformation of two interacting cells is of major interest as conformational changes can be the precursor of important biological phenomena [1]. Presently, the best-resolved marker-free 3D images of cells have been obtained using computational tomographic diffrac-tion microscopy (TDM). It consists in reconstructing digitally the sample contrast from a stack of holograms obtained by inter-ferometry under different illuminations (usually provided by a monochromatic collimated beam with varying angles) [2]. With such a data set, it is possible to form a 3D image with a resolution typically twice better than the standard microscopy techniques. However, most of the studies in TDM have been performed with setups in transmission [2-8], which eases the volume reconstruction of the sample, but ends up with an axial resolution remaining at least 3 times worse than the transverse one. As a result, 3D conformation changes at cell membranes or interfaces in the wavelength range cannot be resolved. To image the cell contour in 3D, the reflection geometry, which is highly sensitive to reflections from interfaces but not to slowly varying volume inhomogeneities, may be more appropriate [9]. In biomedical imaging, this geometry is mainly encountered in optical coherence tomography (OCT) [10], where the axial resolution, of at best one micron, remains insufficient to detect sub-micrometer axial deformation such as those encountered in lymphocyte activation [1]. In this context, reflection TDM, with its theoretical twice better resolution [11], seems an interesting solution. So far, this computational approach has been applied to image thin structures deposited or etched on a reflective sub-strate [12-16], and to obtain 2D reflectance images of cells [17]. Combined with broadband illumination, it was also used to image reflective targets under a thick diffusive layer [18,19] and biological tissues at different depths [20,21], the digital reconstruction allowing an efficient removal of the multiple scattering and aberrations deteriorating the images of standard OCT. In this Letter, we apply reflection TDM to the 3D imaging of cells. We show how a TDM setup can be used as a synthetic confocal microscope, and we take advantage of this computational approach to correct the focus aberrations induced by the use of a high numerical aperture oil-immersion objective (NA = 1.49). We compare reflection and transmission TDM on simulated data and provide experimental images of calibrated and biological samples. TDM permits one to retrieve the 3D map of refractive index of a sample from the measurement of its scattered field under various illumination angles, using a numerical reconstruction procedure. A sensitivity to refractive index contrasts below 0.01 is typically attained [5-8]. Usually the field is measured in a plane conjugated to the sample, and then transferred to the far-field (Fourier space) with a 2D discrete Fourier transform to ease the data treatment steps. The simplest link between the sample refractive index map and the scattered field is obtained under the Born approximation , typically valid for samples with weak refractive index contrast [2]. In this case, in the scalar approximation, the field scattered in far-field along wave vector k s for an illumination plane wave along wave vector k i is given by E s (k s , k i) ∝˜ ε(k s − k i), (1) 0146-9592/20/133721-04 Journal