Low-dose Spectral Computed Tomography for Measuring Breast Tissue Composition

 

Spectral CT

 

Among the most frequently diagnosed cancers, invasive breast cancer was expected in 2011 to be diagnosed at an estimated 230,480 new cases in U.S. women with about 39,520 women expected to die from the disease Currently, the standard imaging modality for breast cancer screening is X-ray projection mammography. Despite mammography’s impressive advantages in detection performance, imaging time, and cost-effectiveness, the imaging community widely recognizes its limitations, inherited from the nature of the 2D projection technique. One of mammography’s fundamental challenges is the superimposition of the breast anatomy on a 2D projection image, which results in reduced contrast resolution. Further, the overlap of the normal breast parenchyma may obscure tumor identification, exacerbated when dense breasts are imaged. Such limitations have increased the interest in improving the sensitivity of breast imaging techniques, especially for dense breasts that may be seen among women of all ages but are more common among younger women.


An ideal breast imaging modality should offer full 3D and dynamic information with high contrast and spatial resolution to detect soft tissue lesions and microcalcifications. The average dose delivered to a patient should be less than the current standard of 6 mGy. The scanning time and patient positioning should also be considered in order to improve patient comfort. The recent development of improved solid state detector technologies and fast readout electronics have led to the consideration of breast CT based on energy-resolved photon-counting detectors as the best candidate. Unlike conventional flat panel detectors, which generally use the integrated electrical charge induced by X-ray radiation, photon-counting detectors are able to use multiple thresholds to count individual photons within given energy windows by sorting them according to their energy levels. Thus, a proper selection of the background threshold can effectively reject electronic noise, which otherwise cannot be easily removed from charge-integrating detectors. The energy discrimination ability also offers the potential to vary the contributions to the overall signal of photons in different energy bins. In place of using a suboptimal energy-proportional weighting function, as flat panel detectors do, photon-counting detectors can weight photons differently to produce an optimal contrast-to-noise ratio. Optimal photon weighting further improves the signal-to-noise ratio (SNR). It has been shown, for example, that the SNR of breast CT can be improved by 30 to 90% using a photon-counting detector with photon energy-weighting capabilities. In comparison to standard CT, a photon-counting breast CT could improve SNR by 30%, resulting in a 40% decrease in patient dose. In addition, photon-counting detectors with high energy resolution can be used for accurate material decomposition by setting multiple energy thresholds which eliminate the spectral overlap in conventional dual energy CT.
We propose to design and develop the first clinical multi-slit, multi-slice (MSMS) dedicated breast spectral CT system based on an energy-resolved photon-counting detector. In addition, we intend to focus on identifying and quantifying breast composition in terms of the volumetric fractions of water, lipid, and protein contents. The proposed three-compartment model offers several potential means of enhancing the clinical utility of breast images.


First, the proposed model would allow lesions to be characterized according to their composition. The traditional procedure for cancer assessment of suspicious non-calcified masses in mammography images employs several descriptive factors. Despite the technical advances in mammography, the positive predictive value of biopsy may be as low as 20%, and benign findings account for a large portion of biopsy results. It is thus critical to develop new image-based techniques to improve mammography’s predictive power. A recent report suggests that, in addition to irregular mass shape, speculated mass margin and patient age, high mass density increases the likelihood of malignancy. However, high mass density by itself is not sufficiently accurate to avert the need for a biopsy. On the other hand, if a mass’s quantitative composition could be obtained with dedicated breast CT measurements, predictive capability might be improved. Several reports have studied the X-ray attenuation properties of normal and neoplastic breast tissues. It has been suggested that differences in linear attenuation coefficients between malignant and normal breast tissues can be significant below 31 keV, especially when different tissue types from the same patient are compared. Since breast tissue is composed primarily of water, lipid, and protein, the discrepancy in linear attenuation coefficients indicates that the chemical composition of malignant and normal tissues can differ significantly. It has been reported that malignant tissues have a significantly higher water fraction compared to normal tissues. Additionally, lipid content can be used to identify certain benign lesions. These reports indicate that sensitivity and specificity may be improved if the signals from water, lipid and protein can be isolated to better characterize a lesion according to its composition. The proposed system can then be used for improved diagnosis of high-risk patients, in place of current clinical imaging modalities, such as ultrasound and MR. The detector’s energy resolution gives it the potential to improve the identification of malignant lesions and thus reduce the number of biopsies needed for suspicious lesions as well as the number of benign biopsy findings.
Second, the proposed model would provide quantified metrics to stratify women according to breast cancer risk. The current standard of care for breast density evaluation involves visual assessment of mammograms using the four-category Breast Imaging Reporting and Data System (BI-RADS). This subjective classification scheme is limited by its considerable intra- and inter-reader variability. Moreover, the current two-compartment model, which assumes that breast is composed of fibroglandular and adipose tissues, suffers large uncertainty from system calibration due to wide variations in the chemical composition of fibroglandular and adipose tissue. Extending the traditional model into a three-compartment model which decomposes breast tissue into its fundamental water, lipid, and protein contents enables quantitative categorization according to breast composition. Thus, the water, lipid and protein decomposition has the potential to provide cancer risk evaluation superior to that of mammographic breast density.


Finally, a previous report has also suggested that tissue’s water content can be affected by pathology. It would be of great importance to study the correlation between changes in the surrounding tissue’s composition, such as increased water content, and the tumorgenesis process. Mapping the water fraction in tissue may provide additional relevant information to help characterize tumor progression and response.