Phase-sensitive techniques extend the possibilities of X-ray imaging by exploiting both the phase and attenuation information [1]. They are of particular interest for medical applications as they enable objects composed of low Z materials to be explored with improved contrast and reduced dose compared to standard absorption radiography. Two X-ray phase-contrast imaging techniques were investigated at ID19, for a future application to mammography: the "propagation-based" (PB) and the "analyser-based" (AB) phase-contrast techniques. The first simply uses the propagation (Fresnel diffraction) of the partially coherent X-ray beam over large distances. The latter uses an analyser crystal, which is down stream of the sample and which acts as an angular filter for X-rays refracted by the sample.

In first approximation, the PB technique is sensitive to the two-dimensional Laplacian of the phase in a plane perpendicular to the incident beam whereas the AB technique is sensitive to the one dimensional gradient of the phase parallel to the plane of diffraction. Therefore, the PB technique produces contrast for objects with density or thickness variations, which are not linear such as micro-calcifications or fibrous structures. The AB technique is better adapted to rendering the low spatial frequencies in the object such as tumour mass lesions, but only in one direction.

Extraction of quantitative information was performed on both analyser- and propagation-based images. A local and statistical method has been proposed to disentangle the absorption, scattering and refraction information from AB images [2] and, for the first time with this approach, a quantitative phase map could be retrieved from the refraction component. The holographic phase retrieval method based on the distance variation was applied on PB images.

Figure 146 shows the astonishing contrast that can be observed on a phase map of a biopsy containing human breast tissues, as compared to the image obtained in simple absorption radiography.


Fig. 146: Images of a 13 mm diameter biopsy sample containing human breast tissue, (a) Absorption image recorded at 0.03 m (sample-to-detector distance). (b) Phase map retrieved from 4 propagation-based images recorded at 0.03 m, 1 m, 4.3 m and 8.8 m. (c) Phase map retrieved from five analyser-based images, two acquired at the tails at ± FWHM, two acquired at the flanks at ± 1/2 FWHM and one acquired at the peak of the analyser crystal rocking curve. Energy: 25 keV, exposure time: 1 s, pixel size: 7.5 µm.


The determination of the phase for different orientations of the sample serves as input for three-dimensional imaging to reconstruct variations of the mass density Dr inside the object. Indeed, up to a proportionality factor the phase maps are identical to projections of Dr. Figure 147 shows a computed tomography (CT) slice of the 13 mm diameter biopsy sample reconstructed from 1200 projections in absorption mode (a), mapping the linear attenuation coefficient in the tissues, in propagation based mode (b) and in analyser based mode (c), mapping the changes in mass density of the tissues.


Fig. 147: Tomography slices of a 13 mm diameter biopsy sample. (a) Map of the linear attenuation coefficient in the breast tissues. (b) and (c) Maps of the variations of the mass density in the breast tissues, the colour-bar is inversed with respect to the absorption case. In the case of the CT (b), the set of projections (maps of the phase), was obtained after processing of propagation based images, and after processing of analyser based images, in the case of the CT (c). The magnified portion in (b) clearly reveals the collagen and fatty cells.


Although the density map obtained with the AB technique (c) is much more noisy than that obtained with the PB technique (b), they both show a net improvement of contrast, revealing details such as the presence of collagen, that are not perceptible in the absorption CT (a).

Density CT provides clear information, and the tomography slices are qualitatively similar to a very large number of histological sections, at the same spatial resolution. This could represent an advantage for the characterisation of biopsies, enabling the bulk of the sample to be rapidly examined and avoiding the constraints and artefacts related to the sample preparation for histology. The comparison of the two phase sensitive techniques provides furthermore valuable information for future application in mammography [3].

[1] R. Fitzgerarld, Physics Today, 53, 7 (2000).
[2] E. Pagot, P. Cloetens, S. Fiedler, A. Bravin, P. Coan, J. Baruchel, J. Härtwig and W. Thomlinson, Appl. Phys. Lett. 82, 20 (2003).
[3] E. Pagot, S. Fiedler, P. Cloetens, A. Bravin, P. Coan, J. Baruchel, K. Fezzaa and J. Härtwig, Phys. Med and Biol., (2005), in press.

E. Pagot (a), P. Cloetens (a), S. Fiedler (a), A. Bravin (a), P. Coan (a), K. Fezzaa (b), J. Baruchel (a), J. Keyriläyinen (c).
(a) ESRF
(b) APS
(c) Helsinki University Central Hospital, HUCH (Finland)