The classical approach for determining stellar angular diameters is to use interferometry and to measure fringe visibilities. Indeed, in the case of a source having a diameter larger than typically λ/6B, B being the interferometer's baseline and λ the wavelength of observation, the fringe contrast decreases. Similarly, it is possible to perform angular diameter determinations by measuring the stellar leakage from a coronagraphic device or a nulling interferometer. However, all coronagraphic devices (including those using nulling interferometry) are very sensitive to pointing errors and to the size of the source, two factors with significant impact on the rejection efficiency. In this work, we present an innovative idea for measuring stellar diameter variations, combining coronagraphy together with interferometry. We demonstrate that, using coronagraphic nulling statistics, it is possible to measure such variations for angular diameters down to ≈λ/40B with 1σ error-bars as low as ≈λ/1500B. For that purpose, we use a coronagraphic implementation on a two-aperture interferometer, a configuration that significantly increases the precision of stellar diameter measurements. Such a design offers large possibilities regarding the stellar diameter measurement of Cepheids or Mira stars, at a 60-80 μas level. We report on a simulation of a measurement applied to a typical Cepheid case, using the VLTI-UT interferometer on Paranal.