Volumetric High-Resolution X-Ray Phase-Contrast Virtual Histology of Breast Specimens With a Compact Laboratory System

Abstract

The assessment of margin involvement is a fundamental task in breast conserving surgery to prevent recurrences and reoperations. It is usually performed through histology, which makes the process time consuming and can prevent the complete volumetric analysis of large specimens. X-ray phase contrast tomography combines high resolution, sufficient penetration depth and high soft tissue contrast, and can therefore provide a potential solution to this problem. In this work, we used a high-resolution implementation of the edge illumination X-ray phase contrast tomography based on "pixel-skipping" X-ray masks and sample dithering, to provide high definition virtual slices of breast specimens. The scanner was originally designed for intra-operative applications in which short scanning times were prioritised over spatial resolution; however, thanks to the versatility of edge illumination, high-resolution capabilities can be obtained with the same system simply by swapping x-ray masks without this imposing a reduction in the available field of view. This makes possible an improved visibility of fine tissue strands, enabling a direct comparison of selected CT slices with histology, and providing a tool to identify suspect features in large specimens before slicing. Combined with our previous results on fast specimen scanning, this works paves the way for the design of a multi-resolution EI scanner providing intra-operative capabilities as well as serving as a digital pathology system.

Fig. 1

Panel (a) and (b) show the experimental LSF and a schematic view of the corresponding EI system for the non-skipped and skipped mask configurations, respectively. For both panels, the corresponding experimental and smoothed ESF (the derivative of which yields the LSF) is also reported in the insets. A fit of both LSF profiles is also shown. In the skipped mask case, 8 dithering steps have been acquired and recombined prior to the ESF estimation.

Fig. 2

Panel (a), (c) and (b), (d) show a CT slice from two specimens acquired with both low resolution and high resolution configurations, respectively. Red arrows in panel (d) point at fine features invisible or barely visible in the low resolution scan. Scale bar is 1 mm for all the panels.

Fig. 3

Panel (a) shows a CT slice of a fixed breast specimen obtained with the skipped sample mask. Panel (b) to (c) show a comparison between ROIs extracted from the high-resolution CT scan obtained with skipped sample masks (left hand side) and the low-resolution scan acquired with the non-skipped sample mask (right hand side). Red arrows in the high-resolution ROIs points at fine strands of tissue that are invisible in the low-resolution scan. Scale bar is 500 μm.

Fig. 4. Comparison between XPCT and H&E staining.

Panels (a) and (b) show matched high-resolution CT and histology slices, respectively. Red and blue arrows in panel (a) point at regions with different density as observed in the CT slice, with the zoomed-in regions in panel (b) highlighting the different cellular structure of the same areas. Panels (c) and (d) show a zoomed in comparison between XPCT and histology slices. Red, blue and yellow arrows point at clusters of immune cells, a milk duct and a thin strand of tissue, respectively. XPCT and histology slices are overlapped in panel (e). A similar comparison is reported in panels (f) and (g) where yellow arrows point at areas of response to chemotherapy, which appear with a lower density. The red arrow points at a region of a residual infiltrating ductal carcinoma nodule.