There have recently been several studies published involving terahertz (THz) imaging of frozen biomedical samples. In this paper, we investigate the effects of the freeze-thaw cycle on THz properties of porcine muscle and fat samples. For ordinary freezing, there was a significant change in the THz properties after thawing for muscle tissue but not for fat tissue. However, if snap-freezing was combined with fast-thawing instead of ordinary freezing and ordinary thawing, then the freeze-thaw hysteresis was removed.
© 2016 Optical Society of America
In recent years there has been increasing interest to use terahertz time-domain spectroscopy (THz-TDS) in biomedical investigations . Several studies have proposed interesting uses of THz waves such as to delineate cancer margins [2–4], monitor the water content in living plants [5, 6], and detect dental defects [7, 8].
THz light is sensitive to subtle changes in tissue water content and structure  that can be caused by diseases such as cancer which suggests that terahertz imaging is potentially a very useful tool for quantitative investigation. At present, most THz research is done ex vivo but freshly excised samples may dehydrate and deteriorate. Several sample handling methods have been devised to avoid complications associated with sample deterioration. For instance, prior to THz measurement one can immerse the sample into a culture medium  or embed it in gelatin  to prevent sample dehydration, and measure as soon as possible. Such methods require an isotonic medium to balance the osmotic pressure. Alternatively, removing the water content in the ex vivo tissue by lyophilization  or formalin fixing  may avoid issues caused by tissue hydration. Lyophilized or fixed samples are more rigid than fresh samples and are more robust. Furthermore, studies have shown that measuring ex vivo samples below 0°C reduces water absorption and improves the penetration depth. For example, Hoshina et al  measured frozen porcine samples at −33°C and the absorption of the tissue was reduced dramatically but there was still enough contrast to distinguish between the striated muscle and the adipose tissue on the sample. Sim et al  used dry ice to freeze excised oral cancer tissue to −20°C, achieving improved THz image contrast between normal and cancerous tissue. In this work we show that the THz properties of tissues are affected by the methods used to freeze and thaw the samples and explain how this in turn will affect the image contrast.
We test two methods to freeze and thaw muscle and fat porcine samples: the first way was to freeze samples at −20°C for three days and then thaw at room temperature. The second way was to snap-freeze the samples with liquid nitrogen and store them in a −80°C freezer for three days and then thaw in a water bath. Both fresh and thawed tissues were imaged with THz radiation in reflection geometry for comparison.
Samples were manually cut sequentially before THz measurement. Each sample was about 1 cm (width) x 1.5 cm (height) x 0.3 cm (thickness). To identify which side should be measured and to help with image registration, one corner of the sample was cut off. We used 3D printing to make a mold so as to cut the samples into slices of uniform thickness. The subcutaneous fat samples we used were stiffer than the muscle samples so it was easier to cut them uniformly.
We used a commercially available time-domain THz spectrometer (TERA-K15, Menlo Systems GmbH, Martinsried, Germany) to measure the samples in reflection geometry with a fixed incident angle of 30° (other angles could be used if there is enough space to fit in the required collimating/focusing lenses). Since the samples have high attenuation, reflection geometry is preferred. Each sample was placed flat on the quartz window of the THz-TDS system and gentle pressure was applied to remove any air gaps. The sample was then raster scanned resulting in an imaging time of less than 2 minutes.
The sample was weighed immediately after the THz measurement. To slow-freeze the sample we first wrapped the sample with cling film and foil, and then put it in to a −20°C refrigerator for three days before thawing. Whereas for snap-freezing we first wrapped the sample with cling film and cooled it to 4°C by placing the wrapped sample into a refrigerator. Then the sample together with cling film was directly immersed in liquid nitrogen for 90 seconds. After 90 seconds there was no boiling observed near the sample, indicating that the sample was at the same temperature as the liquid nitrogen, the sample was then immediately wrapped with foil and stored in a −80°C freezer for 3 days.
Slow-freeze samples were thawed at an ambient room temperature of 18°C and the moisture released by the sample was dabbed off during the process. Fat samples took about 5 minutes to thaw but muscle samples took about 10-15 minutes because they had a higher specific heat capacity.
To speed up the thaw rate and prevent water crystallization  in snap frozen samples, a 37°C water bath was used. Each sample was sealed in an air tight freezing bag and placed in the water bath for 1 minute; we will call this process snap-thawing for clarity.
Before THz measurement, each sample was measured by a thermometer to ensure it had reached ambient room temperature (18°C). We also measured the sample weight again to check for any weight change after the freeze-thaw cycle.
For each tissue type, over 400 data points were extracted from 6 samples. Samples used on the same day were cut from the same piece of meat. Each sample had about 70 useable data points, depending on the sample size and contact with the window. Previous work by our group details our calculation methods to extract the refractive index and absorption coefficient from the data. We used the reflection from the lower surface of the quartz window to account for optical drift, fluctuations in the THz signal, thickness variation in the quartz window  and we used a measurement of water to calculate unwanted reflections to be removed as a baseline .
Slow-freeze, slow-thaw and snap-freeze, snap-thaw data
The average weight loss was 2.36% for muscle samples after the slow-freeze, slow-thaw cycle. All three muscle samples showed a significant decrease in the refractive index. Figure 1 shows that on average, the refractive index decreased by 0.082 at 0.4 THz and the difference is similar across the frequency range of 0.1-1.0 THz. No specific spectral features were observed across the frequency range measured; we choose to discuss the results for 0.4 THz because the power output of our THz system is greatest near this frequency. The error bars represent the 95% confidence interval and are so small that they are nearly indistinguishable.
To highlight the edge effects, Fig. 2(a) shows the refractive index at 0.2 THz measured from a fresh muscle tissue sample; the corresponding image of the thawed sample (Fig. 2(b)) is a slightly different shape and colour, indicating that there was some deformation during the slow-freeze, slow-thaw process. All THz images were acquired at room temperature using a single point raster scanning approach  with a spot size around 2 mm, and a pixel size of 1.30 mm (horizontal) by 0.49 mm (vertical). In the horizontal direction, the scan step is larger because the horizontal movement of the X-Y stages limits the image acquisition speed. If the measurement time is long, the sample might deteriorate and affect the reliability of the measurements. It is necessary to balance the acquisition time and the number of the data points. We have minimized the image acquisition time to reduce the effect of tissue variation during image acquisition and a representative amount of data points (around 70) were collected per sample. The regions of interests were segmented by first generating a mask that covers the region with high contrast, and then dilating the mask to cover the entire sample before removing the edge of sample to avoid edge effects. Details of the image segmentation can be found in .
In contrast, there was no significant change in either the refractive index or the absorption coefficient from slow-freeze, slow-thawed fat samples. The average weight loss was 0.31% which was the lowest average from all freeze thaw cycles of both muscle and fat tissues. From the THz images shown in Fig. 2(c)-2(d) we can see that the overall shape was well preserved.
We also measured the slow-freeze muscle samples immediately after removal from the freezer, but the samples were very stiff so made poor contact with the quartz window. Additionally there was some condensation. Despite these issues, we managed to capture the decrease in both the refractive index and absorption coefficient from one of the muscle samples when it was still cold by point scanning. Figure 3 shows the THz properties when the sample is fresh, cold and thawed. The decrease in both the refractive index and absorption coefficient is consistent with results previously reported [13, 14] where the samples were frozen with the imaging window and the temperature was controlled. When the sample was thawed, the refractive index showed significant hysteresis. Figure 3(c) shows the absorption coefficient against the refractive index from 0.2 – 0.8 THz during the freeze-thaw cycle: the cold sample had a lower absorption and a lower refractive index compared to when it was fresh and at room temperature and so the curve shifts toward the origin. The sample was then thawed and the curve shifts back towards its original position, but does not reach it, instead it remains slightly shifted to the left. This leftward shift is due to the refractive index of the thawed sample being statistically significantly lower than when it was fresh.
In contrast, for snap-freezing followed by snap-thawing, as illustrated in Fig. 4(a) and 4(b), there was no significant change in the refractive index or absorption coefficient of the muscle samples from 0.2 to 0.8 THz and the weight loss was only 1.19%. The abrupt change at 0.1 THz and 1.0 THz is likely to be due to the lower power output of the system at these frequencies causing a reduced dynamic range. The average weight loss for the fat samples was 1.35% and the average refractive index showed no significant change from 0.2 THz to 0.9 THz see (Table 1).
Additional data: slow-freeze, snap-thaw and snap-freeze, slow-thaw
In order to find out whether the slow-freeze process and/or the slow-thaw process causes the hysteresis, we performed follow-up experiments: slow-freeze, snap-thaw and snap-freeze, slow-thaw. Both of these combinations did not alter the THz properties of the fat samples. However, hysteresis was observed for both of these combinations for the muscle samples, suggesting that both snap-freeze and snap-thaw processes are needed to protect the muscle tissue.
The change of the refractive index, ∆n, caused by the four combinations of freeze-thaw cycles is plotted in Fig. 5 along with ∆n of the cold sample: it is clear that the snap-freeze, snap-thaw process introduced the smallest change in the refractive index. Table 1 presents the weight loss and the change in refractive index at 0.4 THz for the four freeze-thaw cycle combinations. Further experiments are required to quantitatively determine the relationship between the weight loss of the sample and the change in refractive index. However from this study we can see that it is not primarily the overall weight loss of the sample that affects the refractive index, but the processes that the sample has endured. For instance for the slow-freeze muscle tissue experiments, the weight change was similar for the snap-thaw process (2.37%) compared to the slow-thaw process (2.36%) but the refractive index change for the snap-thaw process was noticeably lower (0.073 compared to 0.082). This suggests that the damage to the cell structure from the slow-thawing significantly contributes to the refractive index change. This is likely to be because the cell damage facilitates more water loss. Indeed, if we compare the snap-freeze, slow-thaw with the snap-freeze, snap-thaw for muscle we see that the refractive index change is reduced from 0.066 to zero. This highlights the need to thaw quickly as well as freeze quickly. Whilst we have been able to measure the overall weight loss of the sample, it is difficult to determine whether the weight loss is due to water alone. For instance the greater weight change observed from snap-freeze and snap-thawed fat compared to snap-freeze and slow-thawed fat is probably due to the water bath liquefying some of the fat present within the adipose cells. Similarly, it is also possible that the weight change of the snap-thawed muscle samples is not entirely due to water loss and that other components of the muscle sample are liquefied in the water bath.
Slow-freezing creates ice crystals between cells , leading to irreversible damage of the extracellular matrix and eventually moisture loss when thawed. Muscle samples have more water content (~70%) than fat samples (~30%), therefore their water loss is more pronounced during the slow-freeze, slow-thaw process. Furthermore, ice crystals cause structural changes: we observed the texture of the thawed muscle samples was different from the fresh ones. We therefore attribute the decrease in the refractive index of the muscle samples primarily to the effects of moisture loss, this loss being determined by the freezing and thawing processes used. Fat samples, on the other hand, showed no significant change after the slow-freeze, slow-thaw cycle; the weight loss was low and the tissue texture was preserved.
The results of the slow-freeze, slow-thaw experiments show that biomedical tissues with different water content respond differently. Slow-freezing can preserve biomedical tissues with low water content (such as fat) for THz measurement but those with higher water content (such as muscle tissue) will lose more water, and the refractive index will be reduced significantly. Thus, the slow-freeze, slow-thaw cycle will reduce the THz contrast in cases such as muscle-fat comparisons as the refractive index of the muscle will become lower and therefore closer to that of the fat. However the contrast will be improved in cases where the biomedical material of low water content has a higher refractive index such as enamel (n = 3 at 1 THz ): in such cases, after the freeze-thaw process, the muscle will have a lower refractive index and so the difference between the muscle and enamel will be greater. Thus, for inhomogeneous samples, it will depend on the composition of the inhomogeneities as to whether the contrast will be reduced or enhanced.
Snap-freezing is a commonly used technique in both the food industry and for biological sample preservation because it preserves moisture content and structural integrity. Both the number and the size of the ice crystals formed are reduced in snap frozen samples . In our test, the weight change of snap-frozen, snap-thawed muscle tissues (1.19%) was lower than their slow-frozen, slow-thawed counterparts (2.36%), which is in line with expectation. Surprisingly, the weight change of snap-frozen, snap-thawed fat tissues (1.35%) was greater than that of slow-frozen, slow-thawed fat tissues (0.31%). We suspect that during the water bath some fat liquefies and is released from the sample, causing a greater weight change compared to slow-thawed fat samples. Despite the weight change, the THz properties of snap frozen fat samples were consistent.
There are a number of factors that affect the speed of snap-freezing, including the ratio of sample volume to sample surface area, the initial temperature of the sample, the heat conductance of the sample, and the temperature of the cold source. Most biological samples are poor heat conductors. Bigger samples freeze more slowly from the inside and are therefore more likely to form ice crystals, which may expand and crack the sample. Fortunately, snap-freezing worked very well on the samples we used. To reduce the likelihood of sample cracking, we also cooled the samples to 4°C before snap-freezing, thereby reducing the temperature difference between the frozen shell and the inner tissue. Furthermore, to avoid the Leidenfrost effect  it was necessary to move the sample in the liquid nitrogen to improve the cold conduction. Last but not least, it is equally important to thaw a sample quickly so as to reduce the chance of water crystallization. With proper protocol, snap-freezing followed by snap-thawing could preserve both types of tissues for THz measurement.
In summary, our study has demonstrated that both the slow-freeze and slow-thaw cycles can change the THz properties of tissues and thus cause hysteresis in the freeze-thaw cycle. Tissues with higher water content such as skin or muscle tissue will be affected more than those with lower water content such as fatty tissues, and the contrast in THz images of inhomogenous samples will be affected differently depending on the actual composition of the sample. Thus if scientists are going to use THz imaging to investigate fresh biomedical samples at a later date, it would be more meaningful to do this using snap-freezing coupled with snap-thawing rather than ordinary freezing as the slow-freezing and slow-thawing processes may damage the tissue and affect the THz properties. Indeed, THz imaging is very sensitive to changes in biomedical tissues and is therefore potentially able to detect deterioration in samples if the freezing and/or thawing process has not been performed adequately.
This work was partially supported by the Research Grants Council of Hong Kong (Projects 415313 and 14205514).
We would like to thank Dr. Christopher Gilliam for his advice in image processing and Ms. Mei Yi, Melanie Wong for her help in sample preparation.
References and links
1. S. Fan, Y. He, B. S. Ung, and E. Pickwell-MacPherson, “The growth of biomedical terahertz research,” J. Phys. D Appl. Phys. 47(37), 374009 (2014). [CrossRef]
2. R. M. Woodward, V. P. Wallace, R. J. Pye, B. E. Cole, D. D. Arnone, E. H. Linfield, and M. Pepper, “Terahertz pulse imaging of ex vivo basal cell carcinoma,” J. Invest. Dermatol. 120(1), 72–78 (2003). [CrossRef] [PubMed]
3. V. P. Wallace, A. J. Fitzgerald, S. Shankar, N. Flanagan, R. Pye, J. Cluff, and D. D. Arnone, “Terahertz pulsed imaging of basal cell carcinoma ex vivo and in vivo,” Br. J. Dermatol. 151(2), 424–432 (2004). [CrossRef] [PubMed]
4. V. P. Wallace, A. J. Fitzgerald, E. Pickwell, R. J. Pye, P. F. Taday, N. Flanagan, and T. Ha, “Terahertz Pulsed Spectroscopy of Human Basal Cell Carcinoma,” Appl. Spectrosc. 60(10), 1127–1133 (2006). [CrossRef] [PubMed]
5. R. H. J. Daniel, M. Mittleman, and M. C. Nuss, “T-ray imaging,” IEEE J. Sel. Top. Quantum Electron. 2, 679–692 (1996).
6. E. Castro-Camus, M. Palomar, and A. A. Covarrubias, “Leaf water dynamics of Arabidopsis thaliana monitored in-vivo using terahertz time-domain spectroscopy,” Sci. Rep. 3, 2910 (2013). [CrossRef] [PubMed]
8. E. Pickwell, V. P. Wallace, B. E. Cole, S. Ali, C. Longbottom, R. J. Lynch, and M. Pepper, “A comparison of terahertz pulsed imaging with transmission microradiography for depth measurement of enamel demineralisation in vitro,” Caries Res. 41(1), 49–55 (2006). [CrossRef] [PubMed]
9. S. Sy, S. Huang, Y. X. Wang, J. Yu, A. T. Ahuja, Y. T. Zhang, and E. Pickwell-MacPherson, “Terahertz spectroscopy of liver cirrhosis: investigating the origin of contrast,” Phys. Med. Biol. 55(24), 7587–7596 (2010). [CrossRef] [PubMed]
10. S. Fan, B. Ung, E. P. Parrott, and E. Pickwell-MacPherson, “Gelatin embedding: a novel way to preserve biological samples for terahertz imaging and spectroscopy,” Phys. Med. Biol. 60(7), 2703–2713 (2015). [CrossRef] [PubMed]
11. G. M. Png, J. W. Choi, B. W. Ng, S. P. Mickan, D. Abbott, and X. C. Zhang, “The impact of hydration changes in fresh bio-tissue on THz spectroscopic measurements,” Phys. Med. Biol. 53(13), 3501–3517 (2008). [CrossRef] [PubMed]
13. H. Hoshina, A. Hayashi, N. Miyoshi, F. Miyamaru, and C. Otani, “Terahertz pulsed imaging of frozen biological tissues,” Appl. Phys. Lett. 94(12), 123901 (2009). [CrossRef]
15. B. Li and D. W. Sun, “Effect of power ultrasound on freezing rate during immersion freezing of potatoes,” J. Food Eng. 55(3), 277–282 (2002). [CrossRef]
16. S. Fan, E. P. J. Parrott, B. S. Y. Ung, and E. Pickwell-MacPherson, “Calibration method to improve the accuracy of THz imaging and spectroscopy in reflection geometry,” Photonics Research 4(3), A29 (2016). [CrossRef]
17. S. Huang, P. C. Ashworth, K. W. Kan, Y. Chen, V. P. Wallace, Y. T. Zhang, and E. Pickwell-MacPherson, “Improved sample characterization in terahertz reflection imaging and spectroscopy,” Opt. Express 17(5), 3848–3854 (2009). [CrossRef] [PubMed]
18. MathWorks, “Detecting a Cell Using Image Segmentation” (2016), retrieved https://www.mathworks.com/help/images/examples/detecting-a-cell-using-image-segmentation.html.
20. H. Kiani and D.-W. Sun, “Water crystallization and its importance to freezing of foods: A review,” Trends Food Sci. Technol. 22(8), 407–426 (2011). [CrossRef]
21. F. L. Curzon, “The Leidenfrost phenomenon,” Am. J. Phys. 46(8), 825 (1978). [CrossRef]