Published online before print March 16, 2006, 10.1148/radiol.2392041315
(Radiology 2006;239:533-540.)
© RSNA, 2006
Terahertz Pulsed Imaging of Human Breast Tumors1
Anthony J. Fitzgerald, PhD,
Vincent P. Wallace, PhD,
Mercedes Jimenez-Linan, MD, PhD, MRCPath,
Lynda Bobrow, MBBCh, FRCPath,
Richard J. Pye, MA, MD, FRCP,
Anand D. Purushotham, MBBS, MD and
Donald D. Arnone, PhD
1 From TeraView, 302-304 Cambridge Science Park, Milton Road, Cambridge, CB4 0WG, England (A.J.F., V.P.W., D.D.A.); Department of Histopathology (M.J., L.B.) and Cambridge Breast Unit (A.D.P.), Addenbrooke's Hospital, Cambridge, England; and BUPA Cambridge Lea Hospital, Cambridge, England (R.J.P.). Received July 28, 2004; revision requested September 29; revision received April 5, 2005; accepted May 9; final version accepted July 27.
Address correspondence to V.P.W. (e-mail: vincent.wallace{at}teraview.com).
 |
ABSTRACT
|
|---|
The feasibility of using terahertz pulsed imaging to map margins of exposed breast tumors was investigated by imaging 22 excised human breast tissue specimens with carcinoma excised from 22 women (mean age, 59 years; range, 39–80 years). The study was approved by the local ethics research committee, and informed consent was obtained from all patients. The size and shape of tumor regions on terahertz images were compared with those identified at histopathologic examination of the imaged section. Two image parameters were investigated: the minimum of the terahertz impulse function and the ratio of the minimum to the maximum of the terahertz impulse function. The correlation coefficient for the tumor area on images compared with that on a photomicrograph of all 22 samples was greater than 0.82 for both parameters. The shape of the tumor regions on terahertz images also correlated well with that on a photomicrograph (median Spearman rank correlation coefficient, 0.69). Findings of this study demonstrate the potential of terahertz pulsed imaging to depict both invasive breast carcinoma and ductal carcinoma in situ under controlled conditions and encourage further studies to determine the sensitivity and specificity of the technique.
© RSNA, 2006
|
INTRODUCTION
|
|---|
The terahertz region of the electromagnetic spectrum lies between the microwave and infrared regions and covers frequencies of 0.1–10 THz (1 THz = 1012 Hz), which corresponds to wavelengths of 3 mm to 30 µm. In the past decade, applications of terahertz radiation have grown with advances in the ability to efficiently and practically generate and detect broadband pulses in this frequency range (1). Terahertz pulsed imaging (TPI; TeraView, Cambridge, England), which uses these broadband pulses, has been developed as a result of the recent advances.
Terahertz radiation has a number of properties that make it a viable medical imaging technique. For example, terahertz wavelengths are longer than infrared and optical radiation, so scattering in biologic tissue is small in comparison (2). The wavelength is sufficiently short that a submillimeter lateral resolution of more than 200 µm at 3 THz is readily achievable with an axial resolution of 40 µm (3). Terahertz radiation is nonionizing; the power levels used do not cause any detrimental effects to dividing human keratinocytes (4) and are many orders of magnitude lower than those in the recommended safety guidelines (5). The technique has a very high signal-to-noise ratio because of efficient elimination of the background noise with coherent time-gated detection (6). Furthermore, the broadband frequency content can be used for spectroscopic studies to probe the intermolecular interactions between biomolecules (7). Terahertz spectroscopy has been used successfully to characterize DNA (8) and proteins (9,10). Terahertz pulsed imaging has previously been used for imaging of basal cell carcinoma, a form of skin cancer, ex vivo in the laboratory (11) and more recently in vivo (12,13).
In breast-conserving surgery, the aim is to excise the tumor with an adequate margin of normal tissue and minimize the amount of normal tissue removed. In cases where the margins are not clear after resection, the long-term recurrence rate is 10%–20%, compared with 2%–8% for cases where the margins are clear (14). At present, the situation is only identified several days after surgery, after histopathologic examination of the excised tissue margins. A second operation may be needed for close or involved margins, which requires additional hospital resources and increased risk of patient morbidity. Thus, there is a clinical need to accurately define the margins of the tumor during surgery, to conserve normal tissue, and to minimize the number of second surgical procedures. The purpose of this study, therefore, was to investigate the feasibility of terahertz pulsed imaging for mapping margins of tumors in freshly excised human breast tissue.
|
MATERIALS AND METHODS
|
|---|
Patients and Specimen Preparation
From July 2003 to December 2004, we studied tissue, on the basis of the time and availability of the imaging system, from 22 nonconsecutive female patients (mean age, 59 years; range, 39–80 years) undergoing either wide local excision or mastectomy. Approval for the study was granted by the local Cambridge Research Ethics Committee. Signed informed consent, agreeing to research on tissue removed at the time of surgery, was obtained from all patients. All samples were fresh at the time of measurement, except for two samples that had been fixed in formalin.
Samples were first inked (M.J., L.B.) according to a standard protocol so that margins could be identified during the examination of the slides. They were then sliced to expose any palpable lesions. If there was no palpable tumor, the specimen was sliced through the suspected regions of the tumor identified from the radiographs of the excised sample. In two cases later found to contain ductal carcinoma in situ (DCIS), there was no palpable tumor and no obvious calcification. These samples were fixed in formalin prior to slicing according to a standard histopathologic procedure. After imaging, all samples were submitted for routine histopathologic examination.
Image Acquisition
All breast specimens were imaged (A.J.F., V.P.W.) by using a TPI scanner (TeraView). The system uses photoconductive methods to generate and detect terahertz pulses in the reflection mode. The terahertz optics, indicated by the dashed box in Figure 1, focus the beam on the quartz window. To record a full-image data set, the entire terahertz optics and therefore the terahertz beam were raster scanned in the x-y plane. The system gave a usable frequency range of 0.1–3 THz, with an average power of approximately 100 nW. A more detailed description of the system is presented in the article by Wallace et al (12).
The breast specimen under investigation was placed on the quartz window. During imaging, gentle pressure was applied to the specimen to ensure good contact with the quartz. The orientation of the sample on the imaging window, the location of inked colors were noted, and the sample was photographed so that the terahertz images could be correlated with histopathologic findings. The image area was typically 20 x 20 mm, consisted of 75 x 75 pixels, and took less than 5 minutes to acquire. A terahertz impulse function was obtained from each raw terahertz waveform by deconvolving the system response, as described previously (15,16).
Data Analysis and Image Parameters
Examples of terahertz impulse functions for normal breast tissue and an invasive ductal carcinoma are shown in Figure 2. The difference in impulse functions shown in this figure indicates that there are substantial differences in the optical properties (refractive index and absorption coefficient) between normal and diseased tissue. Two image parameters were investigated: the minimum of the terahertz impulse function (Emin) and the ratio of Emin to the maximum of the terahertz impulse function (Emin/Emax). Normal tissue has an impulse function with a large positive amplitude, defined as Emax, while a tumor has an impulse function with a large negative amplitude, defined as Emin.
The two parameters were derived from the impulse functions to assign pixel values on the terahertz images. The parameter Emin highlights the tumor but is also sensitive to air gaps between the quartz plate and the breast tissue. The reason for this sensitivity is that air has a very low refractive index of 1.0 compared with approximately 1.6 for adipose tissue and 2.1 for the quartz image window. This results in an impulse function that has a high Emax, owing to the mismatch in the quartz-air refractive indices, and a large Emin, owing to the air-tissue interface. Emin/Emax was less sensitive to the presence of air gaps because of the normalizing factor of the high Emax, while it provided good contrast between the tumor and the surrounding tissue because of the small Emax value in this case.
Evaluation of Terahertz Images and Statistical Analysis
Terahertz images were evaluated by comparing the size and shape of the region of contrast with the size and shape of the tumor region identified at histopathologic examination of the corresponding tissue section. Digital photomicrographs of the histopathologic specimen were orientated by using the inked margins and photographs of the sample on the imaging window to align with the terahertz images. A rectangular region of interest the size of the terahertz image was overlaid on the digital photomicrograph and was positioned to coincide with the imaged section of tissue.
Tumor regions were determined by means of a detailed examination of the microscopic slides (M.J., L.B.). These were then transcribed (A.J.F.) to digital photomicrographs by using custom software (Matlab, version 14; Mathworks, Natick, Mass), and the area of the tumor within the region of interest was calculated.
Regions of contrast representing the tumor on the terahertz images were selected by using the same software (A.J.F.). For Emin, the threshold criterion for inclusion as part of the tumor region was a pixel value greater than one-third of the maximum Emin value on the image. For Emin/Emax, the threshold was one-quarter of the maximum value. The areas of tumor on the images were then calculated for each parameter. A Pearson product moment correlation coefficient was calculated for each parameter to determine the correlation between the areas on the images and on the photomicrographs for all 22 samples (Fig 3). A gradient was calculated for the best-fit line by using least-squares linear regression (Excel 2000; Microsoft, Redmond, Wash).
View larger version (21K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3a: Graphs show the relationship between tumor area on terahertz images and photomicrographs for (a) Emin and (b) Emin/Emax. Straight line is the least-squares regression fit and has gradient of 0.91 ± 0.09 for Emin and 0.90 ± 0.09 for Emin/Emax.
|
|
View larger version (22K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3b: Graphs show the relationship between tumor area on terahertz images and photomicrographs for (a) Emin and (b) Emin/Emax. Straight line is the least-squares regression fit and has gradient of 0.91 ± 0.09 for Emin and 0.90 ± 0.09 for Emin/Emax.
|
|
To compare the shape of the tumor regions seen on the images with that seen on photomicrographs, the pixel numbers within the regions identified as tumor were summed across both x and y directions. Thus, the two-dimensional tumor shapes on the images and photomicrographs were reduced to two one-dimensional line profiles as illustrated in Figure 4. We used the Spearman rank correlation (Matlab, Version 14.0) in preference to a Pearson product-moment correlation in this case, because the nonparametric rank allowed comparison of the whole dimensionless profiles and gave a better indication of a relationship despite any nonlinear influences that may arise between the image and histopathologic finding. The Spearman rank correlation calculation of the x and y profiles with the equivalent photomicrograph profiles requires equal numbers of sample points, so the profiles from the photomicrographs were decimated to match the size of the profiles for the images. To facilitate comparison of the shape profile correlations between image parameters and across different cases, the x and y correlations for each image parameter were reduced to a single representative statistic by determining the median of the x and y correlations.
|
RESULTS
|
|---|
Table 1 gives details of the histopathologic findings for each of the 22 specimens along with the area determined for the terahertz images and photomicrographs. Two invasive lobular carcinomas, 14 invasive ductal carcinomas, three mixed invasive ductal and lobular carcinomas, and two cases of pure DCIS were identified. One dense radial scar was also identified. Figures 5–8 show examples of the different histopathologic findings. For consistency, in the color maps highlighting tumor on the images, pixel values representing Emin images were plotted as –1 · Emin.
View this table:
[in this window]
[in a new window]
|
Table 1. Main Pathologic Details and Comparison of Tumor Area on Terahertz Images with Area on the Photomicrograph
|
|
View larger version (191K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5a: Case 21. (a) Photomicrograph of lobular carcinoma. Dashed box indicates region of tissue that was imaged with terahertz imaging. Terahertz images of (b) Emin and (c) Emin/Emax.
|
|
View larger version (109K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5b: Case 21. (a) Photomicrograph of lobular carcinoma. Dashed box indicates region of tissue that was imaged with terahertz imaging. Terahertz images of (b) Emin and (c) Emin/Emax.
|
|
View larger version (123K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5c: Case 21. (a) Photomicrograph of lobular carcinoma. Dashed box indicates region of tissue that was imaged with terahertz imaging. Terahertz images of (b) Emin and (c) Emin/Emax.
|
|
View larger version (101K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6a: Case 7. (a) Photomicrograph of specimen with invasive ductal carcinoma with lobular features. Dashed area indicates region of tissue that was imaged with terahertz imaging. Terahertz images of (b) Emin and (c) Emin/Emax. Two terahertz images of overlapping regions were acquired and have been tiled together. Bright region dominating b was caused by an air gap.
|
|
View larger version (106K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6b: Case 7. (a) Photomicrograph of specimen with invasive ductal carcinoma with lobular features. Dashed area indicates region of tissue that was imaged with terahertz imaging. Terahertz images of (b) Emin and (c) Emin/Emax. Two terahertz images of overlapping regions were acquired and have been tiled together. Bright region dominating b was caused by an air gap.
|
|
View larger version (96K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6c: Case 7. (a) Photomicrograph of specimen with invasive ductal carcinoma with lobular features. Dashed area indicates region of tissue that was imaged with terahertz imaging. Terahertz images of (b) Emin and (c) Emin/Emax. Two terahertz images of overlapping regions were acquired and have been tiled together. Bright region dominating b was caused by an air gap.
|
|
View larger version (128K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7a: Case 5. (a) Photomicrograph of specimen with pure DCIS and only a microinvasive carcinoma component. Dashed box indicates region of tissue that was imaged with terahertz imaging. Solid box indicates region of tissue containing DCIS with necrosis. Terahertz images of (b) Emin and (c) Emin/Emax. (d) Photomicrograph of magnified region in the solid box in a shows area containing DCIS with necrosis.
|
|
View larger version (107K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7b: Case 5. (a) Photomicrograph of specimen with pure DCIS and only a microinvasive carcinoma component. Dashed box indicates region of tissue that was imaged with terahertz imaging. Solid box indicates region of tissue containing DCIS with necrosis. Terahertz images of (b) Emin and (c) Emin/Emax. (d) Photomicrograph of magnified region in the solid box in a shows area containing DCIS with necrosis.
|
|
View larger version (93K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7c: Case 5. (a) Photomicrograph of specimen with pure DCIS and only a microinvasive carcinoma component. Dashed box indicates region of tissue that was imaged with terahertz imaging. Solid box indicates region of tissue containing DCIS with necrosis. Terahertz images of (b) Emin and (c) Emin/Emax. (d) Photomicrograph of magnified region in the solid box in a shows area containing DCIS with necrosis.
|
|
View larger version (130K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7d: Case 5. (a) Photomicrograph of specimen with pure DCIS and only a microinvasive carcinoma component. Dashed box indicates region of tissue that was imaged with terahertz imaging. Solid box indicates region of tissue containing DCIS with necrosis. Terahertz images of (b) Emin and (c) Emin/Emax. (d) Photomicrograph of magnified region in the solid box in a shows area containing DCIS with necrosis.
|
|
View larger version (107K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8a: Case 8. (a) Photomicrograph of specimen with invasive ductal carcinoma on top left, DCIS on top right, and fibrosis from previous needle biopsy on bottom. Dashed box indicates region of tissue that was imaged with terahertz imaging. Solid box indicates region of tissue containing DCIS with necrosis. Terahertz images of (b) Emin and (c) Emin/Emax. (d) Magnified region in solid box in a shows invasive ductal carcinoma (arrow) and DCIS (arrowhead).
|
|
View larger version (105K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8b: Case 8. (a) Photomicrograph of specimen with invasive ductal carcinoma on top left, DCIS on top right, and fibrosis from previous needle biopsy on bottom. Dashed box indicates region of tissue that was imaged with terahertz imaging. Solid box indicates region of tissue containing DCIS with necrosis. Terahertz images of (b) Emin and (c) Emin/Emax. (d) Magnified region in solid box in a shows invasive ductal carcinoma (arrow) and DCIS (arrowhead).
|
|
View larger version (56K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8c: Case 8. (a) Photomicrograph of specimen with invasive ductal carcinoma on top left, DCIS on top right, and fibrosis from previous needle biopsy on bottom. Dashed box indicates region of tissue that was imaged with terahertz imaging. Solid box indicates region of tissue containing DCIS with necrosis. Terahertz images of (b) Emin and (c) Emin/Emax. (d) Magnified region in solid box in a shows invasive ductal carcinoma (arrow) and DCIS (arrowhead).
|
|
View larger version (122K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8d: Case 8. (a) Photomicrograph of specimen with invasive ductal carcinoma on top left, DCIS on top right, and fibrosis from previous needle biopsy on bottom. Dashed box indicates region of tissue that was imaged with terahertz imaging. Solid box indicates region of tissue containing DCIS with necrosis. Terahertz images of (b) Emin and (c) Emin/Emax. (d) Magnified region in solid box in a shows invasive ductal carcinoma (arrow) and DCIS (arrowhead).
|
|
Figure 3a shows the relationship between tumor areas on images formed by using the Emin parameter and those on the photomicrographs for all 22 samples. The Pearson product-moment correlation coefficient calculated for this parameter was 0.89 (P < .001), and the gradient of the best-fit line was 0.91 ± 0.09 (standard deviation). Figure 3b shows the same relationship for the images formed by using the Emin/Emax parameter. The correlation in this case was 0.82 (P < .001), with a best-fit line gradient of 0.90 ± 0.09.
Table 2 contains the results of the Spearman rank correlation comparing the shape of the tumor on terahertz images with that on photomicrographs. In all cases, the number of degrees of freedom was over 50, so the significance of the correlation can be inferred from the correlation coefficient itself. The best correlation was for cases 2 and 11, which had median correlations above 0.84 for Emin and Emin/Emax parameters. The worst correlation was for case 15, with a median of 0.04 for Emin and 0 for Emin/Emax parameter. In summary, the range for the median correlation for all 22 samples for the Emin parameter was 0.04–0.90, with a combined median of 0.71. The Emin parameter performed slightly better than the Emin/Emax parameter, which had a range of 0–0.92 and a combined median of 0.69.
View this table:
[in this window]
[in a new window]
|
Table 2. Spearman Rank Correlation Coefficients Calculated for One-dimensional Line Profiles Comparing the Shape of the Tumor Region on the Terahertz Images and Photomicrographs
|
|
|
DISCUSSION
|
|---|
We have shown in this pilot study that under controlled experimental conditions it is possible to locate, in excised human breast tissue, the margins of invasive and in situ carcinomas on terahertz images, with good comparison with histopathologic findings. Two basic image parameters were investigated, Emin and Emin/Emax. Tumor areas determined on images formed by using the Emin parameter correlated better with areas on photomicrographs than did areas on images formed by using the Emin/Emax parameter. The shape of tumor regions on images formed by using the Emin parameter also correlated slightly better with those on photomicrographs compared with those on images formed by using the Emin/Emax parameter, as shown by the fractionally higher Spearman rank correlation coefficient median.
Overall, correlation of both size and shape of the tumor regions with histopathologic findings was moderately high but not strong. Possible reasons are discussed later along with other study limitations. Several cases had very poor correlation for both area and shape. The area on the image with Emin/Emax parameter for case 17 was substantially different from that on the photomicrograph because of air gaps. The shape comparison for case 15 was very poor. The tumor in this case was diffuse in nature, and variations in the distribution led to very low or negative Spearman rank correlation coefficients.
It was expected that the gradient of the least-squares regression lines for the area comparison would be close to 1. The values of 0.91 ± 0.09 and 0.90 ± 0.09 imply that the threshold criterion for tumor on images is reasonably accurate, although edge effects may be responsible for it being slightly lower than 1. Given that some shrinkage is usually associated with formalin fixation of tissue, the thresholds may need subtle revision if substantial shrinkage is indeed found to occur.
Two cases of pure DCIS were identified in this study. In both cases, the tissue was fixed in formalin prior to imaging, according to standard procedure for tissues, with no palpable tumor or calcifications on the radiograph. Both of these cases still showed contrast, which is consistent with contrast seen on magnetic resonance (MR) images of formalin-fixed biologic tissue (17). Other instances of DCIS, together with invasive carcinoma, were apparent on images of fresh tissue. It is therefore likely that pure DCIS should show contrast on terahertz pulsed images in fresh tissue.
The discrimination between histopathologic findings, including benign tissue, with use of the two basic time-domain parameters was limited. Many other parameters can be derived from impulse functions. Further studies should help identify more specific parameters that are possibly related to pulse width or that use frequency components that may optimally highlight different types of tumor and help differentiate malignant from nonmalignant and normal histopathologic findings.
Contrast mechanisms are not well understood at this stage, but there is clearly a large difference in the refractive index between normal breast tissue and carcinoma. One of the major factors contributing to contrast could be an increase in the vasculature due to the release of growth factors that also lead to rapid cell division and higher cell densities. It is well known on the basis of MR and positron emission tomographic studies that physiologic changes associated with tumor generally lead to increased water content (18–21) and decreased lipid concentration (22) compared with normal tissue. Since liquid water exhibits a strong broad absorption peak centered at 5.4 THz (23,24), which extends down to the frequencies used in this study, terahertz pulsed imaging is extremely sensitive to water content. On the other hand, lipid is a weak absorber of terahertz radiation (25), so the reduction in lipid concentration and increase in water content should lead to strong changes in the terahertz reflection.
Water concentration is not the only likely factor contributing to contrast on terahertz images. Authors of a number of studies have imaged biologic samples that were fixed in formalin, dehydrated, and embedded in wax for histopathologic examination and still found contrast between tumor and the surrounding normal tissue (26–28). This suggests that, in addition to lipid changes, other effects such as increased cell density or the presence of certain proteins may also be responsible for contrast.
One of the limitations of the study was that a direct comparison between areas of tumor on photomicrographs and regions of contrast on the images is impaired because of tissue deformation and the accuracy of the alignment. Tissue was imaged fresh, and because of its elasticity may have stretched slightly under the gentle pressure applied to it during imaging. Further, histopathologic preparation, such as formalin fixation, is known to cause shrinkage and distortion of some tissues. Also, while the photomicrographs show the same plane as the terahertz images, they correspond to only a very thin slice, about 10 µm, and therefore may not be exactly representative of the imaged slice. Finally, the alignment of the photomicrographs was based on photographs of the sample on the image window and observation of colored inks visible on the edge of the tissue slice under microscopic examination of the histopathologic slides. The position therefore could only be estimated; it was not an exact coregistration. One way to improve the comparison would be to add registration markers to the tissue prior to imaging so that they are visible on images and photomicrographs. These markers would allow more accurate alignment and could be used to estimate the amount of tissue deformation due to processing. The statistical comparisons between the tumor size and shape on images and photomicrographs would then be more accurate.
Another limitation was the small number of specimens in this pilot study, which focused on malignant masses. A study is now underway to investigate greater numbers of other histopathologic abnormalities, including benign samples such as fibroadenomas, and more normal ductal and lobular tissue. Further exploration will establish a deeper understanding of the interaction of terahertz radiation with breast tissue. Characterization of the frequency-dependent response of tissue to terahertz radiation with spectroscopic measurements would enable more specific image parameters to be identified and aid in the understanding of contrast mechanisms. The results could be used in computer models that we have developed to simulate the interaction of terahertz radiation with biologic tissue (29,30). These models will enable us to simulate normal tissue and tumor by using their optical properties and to investigate the capabilities of terahertz imaging of tumor in breast tissue to determine whether terahertz pulsed imaging could potentially be exploited to interactively aid surgeons to achieve adequate excision of tumor that is closely involved with margins. This would reduce the need for second operations, potentially improve patient outcome, and save hospitals considerable resources.
Currently there are a number of imaging modalities and techniques being investigated as adjuncts to mammography to aid in the diagnosis of breast cancer or to guide biopsy. These include infrared and optical imaging, thermography, electrical impedance, and MR imaging (31,32). Terahertz pulsed imaging provides information, which is complementary to these techniques, with use of a new region of the electromagnetic spectrum that could improve the overall sensitivity to identifying cancer in exposed breast tissue.
|
ACKNOWLEDGMENTS
|
|---|
The authors thank Sarah Pinder, FRCPath, from the Department of Histopathology at Addenbrookes Hospital Cambridge for her advice and comments on the manuscript.
|
FOOTNOTES
|
|---|
Abbreviations: DCIS = ductal carcinoma in situ • Emin = minimum of the terahertz impulse function • Emin/Emax = ratio of Emin to maximum of the terahertz impulse function
Author contributions: Guarantors of integrity of entire study, A.J.F., V.P.W.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, A.J.F.; experimental studies, A.J.F., V.P.W.; statistical analysis, A.J.F.; and manuscript editing, all authors
R.J.P. is a medical consultant to TeraView.
|
References
|
|---|
- Grischkowsky D, Keiding S, van Exter M, Fattinger C. Far-infrared time-domain spectroscopy with terahertz beams of dielectrics and semiconductors. J Opt Soc Am B Opt Phys 1990;7:2006–2015.
- Han PY, Cho GC, Zhang XC. Time-domain transillumination of biological tissues with terahertz pulses. Opt Lett 2000;25:242–244.[CrossRef][Medline]
- Fitzgerald AJ, Cole BE, Taday PF. Analysis of tablet coating thickness, uniformity and integrity using terahertz pulsed imaging. J Pharm Sci 2005;94:177–183.[CrossRef][Medline]
- Clothier RH, Bourn N. Effects of exposure on human primary keratinocyte differentiation and viability. J Biol Phys 2003;29:179–185.[CrossRef]
- Berry E, Walker GC, Fitzgerald AJ, et al. Do in vivo terahertz imaging systems comply with safety guidelines? J Laser App 2003;15:192–198.
- Mittleman DM, Jacobsen RH, Nuss MC. T-ray imaging. IEEE J Sel Top Quant Electr 1996;2:679–692.[CrossRef]
- Smye SW, Chamberlain JM, Fitzgerald AJ, Berry E. The interaction between terahertz radiation and biological tissue. Phys Med Biol 2001;46:R101–R112.[CrossRef][Medline]
- Brucherseifer M, Nagel M, Haring-Bolivar P, Kurz H. Label-free probing of the binding state of DNA by time-domain terahertz sensing. Appl Phys Lett 2000;77:4049–4051.[CrossRef]
- Markelz AG, Roitberg A, Heilweil EJ. Pulsed terahertz spectroscopy of DNA, bovine serum albumin and collagen between 0.1 and 2.0 Terahertz. Chem Phys Lett 2000;320:42–48.[CrossRef]
- Walther M, Fischer B, Schall M, Helm H, Jepsen UP. Far-infrared vibrational spectra of all-trans, 9-cis, and 13-cis retinal measured by terahertz time-domain spectroscopy. Chem Phys Lett 2000;332:389–395.[CrossRef]
- Woodward RM, Wallace VP, Pye RJ, et al. Terahertz pulse imaging of ex vivo basal cell carcinoma. J Invest Dermatol 2003;120:72–78.[CrossRef][Medline]
- Wallace VP, Taday PF, Fitzgerald AJ. Terahertz pulsed imaging and spectroscopy for biomedical and pharmaceutical applications. Faraday Discuss 2004;126:255–264.[CrossRef][Medline]
- Wallace VP, Fitzgerald AJ, Shankar S, Pye RJ, Arnone DD. Terahertz pulsed imaging of basal cell carcinoma ex vivo and in vivo. Br J Dermatol 2004;151:424–432.[CrossRef][Medline]
- Harris JR, Lippman ME, Morrow M, Osbourne KC, eds. Diseases of the breast. 2nd ed. Philadelphia, Pa: Lippincott Williams & Wilkins, 2004.
- Mittleman DM, Hunsche S, Boivin L, Nuss MC. T-ray tomography. Opt Lett 1997;22:904–906.[Medline]
- Cole BE, Woodward RM, Crawley D. Terahertz imaging and spectroscopy of human skin in vivo. In: Haglund RF, Neev J, Wood RF, eds. Proceedings of SPIE: medical imaging 2001—commercial and biomedical applications of ultrashort pulse lasers: laser plasma generation and diagnostics. Vol 4276. Bellingham, Wash: International Society for Optical Engineering, 2001; 1–10.
- Boulter C, Mulroy S, Webb S, Fleming S, Brindle K, Sandford R. Cardiovascular, skeletal and renal defects in mice with a targeted disruption of the Pkd1 gene. Proc Natl Acad Sci U S A 2001;98:12174–12179.[Abstract/Free Full Text]
- Bruehlmeier M, Roelcke U, Blauenstein P, et al. Measurement of the extracellular space in brain tumors using 76Br-bromide and PET. J Nucl Med 2003;44:1210–1218.[Abstract/Free Full Text]
- Ross KF, Gordon RE. Water in malignant tissue, measured by cell refractometry, and nuclear magnetic resonance. J Microsc 1982;128:7–12.[Medline]
- Chen JH, Avram HE, Crooks LE, Arakawa M, Kaufman L, Briton AC. In vivo relaxation times and hydrogen density at 0.063-4.85 T in rats with implanted mammary adenocarcinomas. Radiology 1992;184:427–434.[Abstract/Free Full Text]
- Rofstad EK, Steinsland E, Kaalhus O, Chang YB, Hovik B, Lyng H. Magnetic resonance imaging of human melanoma xenographs in vivo: proton spin lattice and spin-spin relaxation times versus fractional water content and fraction of necrotic tumour tissue. Int J Radiat Biol 1994;65:387–401.[Medline]
- Jakubowski DB, Cerussi AE, Beilacqua F, et al. Monitoring neoadjuvant chemotherapy in breast cancer using quantitative diffuse optical spectroscopy: a case study. J Biomed Opt 2004;9(1):230–238.[CrossRef][Medline]
- Thrane L, Jacobsen RH, Uhd Jepsen P, Keiding SR. THz reflection spectroscopy of liquid water. Chem Phys Lett 1995;240:330–333.[CrossRef]
- Zelsmann HR. Temperature dependence of the optical constants for liquid H2O and D2O in the far IR region. J Mol Struct 1995;350:95–114.[CrossRef]
- Fitzgerald AJ, Berry B, Zinv'ev NN, et al. Catalogue of human tissue optical properties at terahertz frequencies. J Biol Phys 2003;129:123–128.
- Knobloch P, Schildknecht C, Klience-Ostmann T, et al. Medical terahertz imaging: an investigation of histo-pathological samples. Phys Med Biol 2002;47:3875–3884.[CrossRef][Medline]
- Loffler T, Siebert K, Czasch S, Bauer T, Roskos HG. Visualisation and classification in biomedical terahertz pulsed imaging. Phys Med Biol 2002;47:3847–3852.[CrossRef][Medline]
- Berry E, Handley JW, Fitzgerald AJ, et al. Multispectral classification techniques for terahertz pulsed imaging: an example in histopathology. Med Eng Phys 2004;26:423–430.[CrossRef][Medline]
- Pickwell E, Cole BE, Fitzgerald AJ, Pepper M, Wallace VP. Simulation of terahertz pulse propagation in biological systems. Appl Phys Lett 2004;84:2190–2192.[CrossRef]
- Pickwell E, Cole BE, Fitzgerald AJ, Pepper M, Wallace VP. In vivo study of human skin using pulsed terahertz radiation. Phys Med Biol 2004;49:1595–1607.[CrossRef][Medline]
- Bolan PJ, Meisamy S, Baker EH, et al. In vivo quantification of choline compounds in the breast with 1H MR spectroscopy. Magn Reson Med 2003;50:1134–1143.[CrossRef][Medline]
- Boppart SA, Luo W, Marks DL, Singletary KW. Optical coherence tomography: feasibility for basic research and image-guided surgery of breast cancer. Breast Cancer Res Treat 2004;84:85–97.[CrossRef][Medline]
TOP
ABSTRACT
INTRODUCTION
= profiles from photomicrograph, 
= profiles from Emin image. Photomicrographs contain many more pixels than do terahertz images, so these profiles have been decimated to have same size as the image profiles. All profiles have been scaled by their respective means to allow comparison on the same axes.
