In this issue, Li, Giger et al. ( ) report on a technique that uses fractal analysis of mammographic parenchymal patterns for the assessment of breast cancer risk. Among other interesting observations, the authors of the article conclude that a combined group of BRCA 1/2 gene mutation carriers exhibit a statistically different radiographic fractal texture pattern than a carefully selected low-risk group. Texture analysis of mammographic images has been the subject of many studies for more than twenty years. Texture conveys information about characteristics such as breast density, glandularity and the degree of randomness in the mammographic appearance of tissues. Several studies have reported on the association between increased cancer risk and mammographic density ( ). Increasingly, breast density is regarded as another potential factor that can be used in conjunction with mammographic interpretation and other relevant information to assess breast cancer risk. Other types of texture analysis derived from digitized mammograms and correlation with malignancy or cancer risk have been reported by various research groups ( ). It is only in the past decade that these efforts seem to appear increasingly relevant and with significant promise for clinical utility. The discovery of BRCA1 and BRCA2 gene mutations more than a decade ago and further developments in associating the genotype with risk have added a new consideration in strategies with regard to breast cancer screening. The concept of a blood test for BRCA 1/2 or other mutations seems simple enough but most women are not carriers of these mutations and there is a large variability in occurrence of such mutations in the general population ( ). Hence, the search for alternative measures of risk particularly from easily obtainable patient related information seems to make logical sense.
In this study Li, Giger et al. ( ) report on an improved method of analyzing mammographic parenchymal patterns by using fractal analysis within a region of interest in digitized mammograms. Comparison of the parenchymal patterns of the high risk group of subjects with known BRCA 1/2 gene mutations to a carefully selected group that was considered at low risk demonstrated reasonable discrimination between the two groups. Although a similar trend was previously reported by this research group ( ), the present report demonstrated an improvement in discriminating between the high risk and the low risk groups. The results of this study suggest that more information can be extracted from conventional mammograms than what is currently obtained. At this stage it would be naïve to assume that more advanced mammographic parenchymal texture analysis would have the required sensitivity and specificity to identify known or yet-to-be discovered gene mutations. However, the combined information of parenchymal density, texture and potentially other parameters may be used as an adjunct and can be extremely valuable in screening and in assessing breast cancer risk. The results of this study by Li, Giger et al. ( ) reveal an intriguing trend, but currently we have no knowledge on how gene mutations or other risk factors may affect the texture of the breast parenchyma in a way that may be detectable by fractal analysis or other image analysis techniques. Better understanding of these mechanisms may be derived from advances in genetics and pathology in conjunction with advances in image acquisition and analysis techniques. The present work is representative of an interesting trend where imaging is gradually breaking barriers in predictive medicine and assumes an increasingly important role in medical decisions. These results by Li, Giger et al. ( ) were derived from planar film-screen mammograms because the availability of digital mammography cases as needed for this study was limited. Although texture analysis can be performed very effectively from digitized film, certain inherent characteristics of film such as the non-linearity of response, overexposure or underexposure may obscure other information that otherwise might be available for further analysis. The investigators carefully avoided exposure problems by limiting their analysis to properly exposed mammograms. Digital mammography with its wide linear response to exposure lends itself easily for this type of analysis and we can easily envision that such analysis could be performed directly from a digital mammogram. However, the translation of this technique and its future variants to digital mammography will require further evaluation. Moreover, since digital mammography is represented by various technologies with vastly different physical characteristics, the adaptation of these techniques must be made by considering the physical characteristics with particular attention to potential effects due to pixel size and other parameters that affect spatial resolution. It would be interesting to evaluate the capability of discrimination between the high and low risk groups at various spatial resolution conditions. Moreover, digital mammography would enable easy sampling and analysis of other regions of interest or even the entire breast image. Digital image file format should also be considered; for example, the user may have access only to a processed image (a modified image optimized for viewing) but the unprocessed “raw” image file may be more suitable for this type of analysis. Lossy image compression of mammographic images is currently not performed but as the technology evolves some form of image compression may be practiced in digital mammography. Any of these techniques will alter certain characteristics of the image and therefore, the potential effects should be considered.
The analysis presented here may be adaptable to the emerging digital tomosynthesis techniques ( ), but image artifacts, due to the nature of the image acquisition process and image reconstruction, may present significant challenges. It would be intriguing to evaluate such techniques with tomographic three-dimensional (3-D) data sets as acquired by dedicated computed tomography of the breast ( ), a method that is receiving significant attention recently. However, the spatial resolution of dedicated breast CT is somewhat lower than mammography, but in terms of volume element size, breast CT provides more detailed information, thus paving the way for volumetric analysis.
The work by Li, Giger et al. ( ) leads to some interesting considerations: (i) Mammograms and most likely other images may contain additional information that can be extracted by advanced mathematical analysis techniques. (ii) Information extracted from such analysis may provide for better assessment of cancer risk. (iii) This method of analysis seems worthy of further evaluation with digital mammography, tomographic and 3D imaging techniques. (iv) One important advantage of such technique is associated with the potential ease of obtaining this information from standard mammographic imaging without additional procedures.
The association of structure with a potential for malignancy is generating renewed interest in recent years and we begin to appreciate the potential of tissue pattern analysis as a practical, convenient, and relatively inexpensive risk assessment tool.
The author wishes to thank Srinivasan Vedantham, Ph.D., for valuable discussions on this topic. Disclosure: AK has research collaboration with GE Corporate Research and Development through his institution, Emory University. In 2006, he participated in one of General Electric’s advisory board meetings on breast cancer. The content of this editorial reflects solely the author’s views and it is not associated in any way with General Electric or any other sponsor.
References
1. Li H., Giger M.L., Olopade O.I., Lan L.: Fractal Analysis of Mammographic Parenchymal Patterns in Breast Cancer Risk Assessment. Acad Radiol 2007; pp. 513-521.
2. Boyd N.F., Byng J.W., Jong R.A., et. al.: Quantitative classification of mammographic densities and breast cancer risk: results from the Canadian National Breast Screening Study. J Natl Cancer Inst 1995; 87: pp. 670-675.
3. Boyd N.F., Guo H., Martin L.J., et. al.: Mammographic density and the risk and detection of breast cancer. N Engl J Med 2007; 356: pp. 227-236.
4. Byrne C.: Mammographic density: a breast cancer risk factor or diagnostic indicator?. Acad Radiol 2002; 9: pp. 253-255.
5. Saftlas A.F., Hoover R.N., Brinton L.A., et. al.: Mammographic densities and risk of breast cancer. Cancer 1991; 67: pp. 2833-2838.
6. Byrne C., Schairer C., Wolfe J., et. al.: Mammographic features and breast cancer risk: effects with time, age, and menopause status. J Natl Cancer Inst 1995; 87: pp. 1622-1629.
7. Wolfe J.N.: Breast patterns as an index of risk for developing breast cancer. AJR Am J Roentgenol 1976; 126: pp. 1130-1137.
8. Wolfe J.N., Saftlas A.F., Salane M.: Mammographic parenchymal patterns and quantitative evaluation of mammographic densities: a case-control study. AJR Am J Roentgenol 1987; 148: pp. 1087-1092.
9. Tahoces P.G., Correa J., Souto M., Gomez L., Vidal J.J.: Computer-assisted diagnosis: the classification of mammographic breast parenchymal patterns. Phys Med Biol 1995; 40: pp. 103-117.
10. Malone K.E., Daling J.R., Doody D.R., et. al.: Prevalence and Predictors of BRCA1 and BRCA2 Mutations in a Population-Based Study of Breast Cancer in White and Black American Women Ages 35 to 64 Years. Cancer Res 2006; 66: pp. 8297-8308.
11. Anglian Breast Cancer Study Group: Prevalence and penetrance of BRCA1 and BRCA2 mutations in a population-based series of breast cancer cases. Br J Cancer 2000; 83: pp. 1301-1308.
12. Huo Z., Giger M.L., Olopade O.I., et. al.: Computerized analysis of digitized mammograms of BRCA1 and BRCA2 gene mutation carriers. Radiology 2002; 225: pp. 519-526.
13. Huo Z., Giger M.L., Wolverton D.E., Zhong W., Cumming S., Olopade O.I.: Computerized analysis of mammographic parenchymal patterns for breast cancer risk assessment: feature selection. Med Phys 2000; 27: pp. 4-12.
14. Li H., Giger M.L., Olopade O.I., Margolis A., Lan L.: Computerized texture analysis of mammographic parenchymal patterns of digitized mammograms. Acad Radiol 2005; 12: pp. 863-873.
15. Niklason L.T., Christian B.T., Niklason L.E., et. al.: Digital Tomosynthesis in breast imaging. Radiology 1997; 205: pp. 399-406.
16. Boone J.M., Nelson T.R., Lindfors K.K., Seibert J.A.: Dedicated breast CT: radiation dose and image quality evaluation. Radiology 2001; 221: pp. 657-667.