Preliminary observations of breast tumor collagen

Analysis of breast cancer by small angle X-ray scattering (SAXS)

Small angle X-ray scattering (SAXS) images of normal breast tissue and benign and malignant breast tumour tissues, fixed in formalin, were measured at the momentum transfer range of 0.063 nm(-1) < or = q (= 4pisin(theta/2)/lambda) < or = 2.720 nm(-1). Four intrinsic parameters were extracted from the scattering profiles (1D SAXS image reduced) and, from the combination of these parameters, another three parameters were also created. All parameters, intrinsic and derived, were subject to discriminant analysis, and it was verified that parameters such as the area of diffuse scatter at the momentum transfer range 0.50 < or = q < or = 0.56 nm(-1), the ratio between areas of fifth-order axial and third-order lateral peaks and third-order axial spacing provide the most significant information for diagnosis (p < 0.001). Thus, in this work it was verified that by combining these three parameters it was possible to classify human breast tissues as normal, benign lesion or malignant lesion with a sensitivity of 83% and a specificity of 100%.

X-ray scattering for classifying tissue types associated with breast disease

Collagen types I and III can be characterized at the molecular level (at the tens to hundreds of nanometers scale) using small angle x-ray scattering (SAXS). Although collagen fibril structural parameters at this length scale have shown differences between diseased and nondiseased breast tissues, a comprehensive analysis involving a multitude of features with a large (>50) patient cohort has not previously been investigated. Breast tissue samples were excised from 80 patients presenting with either a breast lump or reduction mammoplasty. From these, invasive carcinoma, benign tissue, and normal parenchyma were analyzed. Parameters related to collagen structure, including longitudinal (axial) and lateral (equatorial) features, polar angle features, total scattering intensity, and tissue heterogeneity effects, were extracted from the SAXS patterns and examined. The amplitude of the third-order axial peak and the total scattering intensity (amorphous scatter) showed the most separation between tissue groups and a classification model using these two parameters demonstrated an accuracy of over 95% between invasive carcinoma and mammoplasty patients. Normal tissue taken from disease-free patients (mammoplasty) and normal tissue taken from patients with presence of disease showed significant differences, suggesting that SAXS may provide different diagnostic information from that of conventional histopathology.

Model fitting in two dimensions to small angle diffraction patterns from soft tissue

In our research programme small angle x-ray scattering (SAXS) is used to provide information on the axial arrangement of collagen molecules as well as data about the state of other components of the extra cellular matrix (ECM) in human tissues. Derivation of parameters to describe and simplify the data is required for much of the SAXS patterns analysis. A method is presented here to achieve function fitting to collagen diffraction peaks along with a representation of the underlying diffuse scatter. A simple model was used which proved reliable in fitting a variety of 2D diffraction patterns. The logarithm of the scatter intensity over the area of the scatter image was taken to reduce the range and improve fitting accuracy. Our model was then used to fit the log data. The model consisted of a radial exponential diffuse scatter component added to a specified number of Gaussian peaks. In 2D the peak model is toroidal, each component being rotated about a common specified centre. Initial search parameters from a 1D averaged sector were supplied to the iterative 2D fitting routine. With the aid of data weighting and basic wavelet filtering, successful and reliable fitting of a specified 2D model to real data is achievable. The process is easily automated. Multiple SAXS patterns can be fitted without operator intervention. As described the model is simple enough to converge rapidly and yet allows image data to be parameterized to a form suitable for extracting the requisite information. The fitting method is flexible enough to be extended to achieve a more comprehensive and complex pattern fitting in two dimensions if this turns out to be necessary. It is our intention to implement orientation distribution functions in the near future by including an angular scaling factor.

A semianalytic model to extract differential linear scattering coefficients of breast tissue from energy dispersive x-ray diffraction measurements

The goal of this work is to develop a technique to measure the x-ray diffraction signals of breast biopsy specimens. A biomedical x-ray diffraction technology capable of measuring such signals may prove to be of diagnostic use to the medical field. Energy dispersive x-ray diffraction measurements coupled with a semianalytical model were used to extract the differential linear scattering coefficients [mus(x)] of breast tissues on absolute scales. The coefficients describe the probabilities of scatter events occuring per unit length of tissue per unit solid angle of detection. They are a function of the momentum transfer argument, x=sin(theta/2)/X, where theta=scatter angle and lambda=incident wavelength. The technique was validated by using a 3 mm diameter 50 kV polychromatic x-ray beam incident on a 5 mm diameter 5 mm thick sample of water. Water was used because good x-ray diffraction data are available in the literature. The scatter profiles from 6 degrees to 15 degrees in increments of 1 degrees were measured with a 3 mm x 3 mm x 2 mm thick cadmium zinc telluride detector. A 2 mm diameter Pb aperture was placed on top of the detector. The target to detector distance was 29 cm and the duration of each measurement was 10 min. Ensemble averages of the results compare well with the gold standard data of A. H. Narten ["X-ray diffraction data on liquid water in the temperature range 4 degrees C-200 degrees C," ORNL Report No. 4578 (1970)]. An average 7.68% difference for which most of the discrepancies can be attributed to the background noise at low angles was obtained. The preliminary measurements of breast tissue are also encouraging.

A preliminary study of breast cancer diagnosis using laboratory based small angle x-ray scattering

Breast tissue collected from tumour samples and normal tissue from bi-lateral mastectomy procedures were examined using small angle x-ray scattering. Previous work has indicated that breast tissue disease diagnosis could be performed using small angle x-ray scattering (SAXS) from a synchrotron radiation source. The technique would be more useful to health services if it could be made to work using a conventional x-ray source. Consistent and reliable differences in x-ray scatter distributions were observed between samples from normal and tumour tissue samples using the laboratory based 'SAXSess' system. Albeit from a small number of samples, a sensitivity of 100% was obtained. This result encourages us to pursue the implementation of SAXS as a laboratory based diagnosis technique.

A reconstruction algorithm for coherent scatter computed tomography based on filtered back-projection

Coherent scatter computed tomography (CSCT) is a reconstructive x-ray imaging technique that yields the spatially resolved coherent-scatter form factor of the investigated object. Reconstruction from coherently scattered x-rays is commonly done using algebraic reconstruction techniques (ART). In this paper, we propose an alternative approach based on filtered back-projection. For the first time, a three-dimensional (3D) filtered back-projection technique using curved 3D back-projection lines is applied to two-dimensional coherent scatter projection data. The proposed algorithm is tested with simulated projection data as well as with projection data acquired with a demonstrator setup similar to a multi-line CT scanner geometry. While yielding comparable image quality as ART reconstruction, the modified 3D filtered back-projection algorithm is about two orders of magnitude faster. In contrast to iterative reconstruction schemes, it has the advantage that subfield-of-view reconstruction becomes feasible. This allows a selective reconstruction of the coherent-scatter form factor for a region of interest. The proposed modified 3D filtered back-projection algorithm is a powerful reconstruction technique to be implemented in a CSCT scanning system. This method gives coherent scatter CT the potential of becoming a competitive modality for medical imaging or nondestructive testing.

X-ray forward-scatter imaging: experimental validation of model

In our research program we have investigated, through modeling and related numerical calculations, the potential use of scattered photons for medical x-ray imaging. In this work, we present an experimental validation of the primary and of the forward-scatter x-ray imaging models. Incident polyenergetic photon beams generated from a conventional rotating anode x-ray tube were used. To compare quantitatively the results between primary and forward-scatter imaging, an ionization chamber was used to record the incident air collision kerma, Kair(c). Plots of contrast (C) and the signal-to-noise ratio (SNR) as a function of the imaging task are presented. We have chosen to make measurements with plastics [polymethyl methacrylate (PMMA), polycarbonate, polystyrene, polyethylene, and nylon] placed at the center of a 15 cm diam spherical water phantom. Good agreement between experiment (expt) and prediction (pred) was obtained for many imaging tasks. For example, to image a 2 cm thick PMMA/polycarbonate combination using an 80 kV beam with the primary photons we obtain Cexpt = 0.01 +/- 0.02, Cpred = 0.008 +/- 0.002, SNRexp/square root Kair(c) = 0.86 +/- 1.6(mJ/kg)(-1/2) and SNRpred/square root K(air)c = 0.51 +/- 0.14(mJ/kg)(-1/2). The values obtained by using the theta = 4 degrees scattered field were Cexpt = 0.26 +/- 0.06, Cpred = 0.19 +/- 0.01, SNRexp/square root Kair(c) = 3.8 +/- 0.8(mJ/ kg)(-1/2), and SNRpred/square root K(air(c) = 3.2 +/- 0.3 (mJ/kg)(-1/2) We have, however, shown that using form factor data from different authors can have a significant effect on the predicted values of C and SNR. The use of our semianalytic expressions for the numbers of transmitted and scattered photons combined with our experimental measurements allowed us to quantify the amount of water contamination in our measurements. Some preliminary results in air with biological materials (liver, muscle, water) are also presented. We are confident that our model can be used as a tool for designing and optimizing an x-ray scatter imaging system.