Acoustic-to-articulatory inversion of speech signals via an analysis-by-synthesis method requires the comparison of natural and synthetic speech spectra either indirectly via formant frequencies, or directly via cepstral coefficients. This paper investigates several strategies of cepstral adaptation (affine transformation of cepstral coefficients, bilinear or piecewise linear frequency warping) when X-ray images of the speaker's vocal tract are available. These images enable the articulatory synthesis of a speech signal which fits the natural signal at best. It is thus possible to investigate the behavior of several cepstral adaptation procedures in order to select the best method, i.e. that which minimizes the deviation between synthetic and natural spectra. Our results show that the affine cepstral adaptation tends to flatten the spectral peaks, i.e. formants. Frequency warping techniques are thus more efficient all the more they can be supplemented by taking into account the spectral tilt.