1. Introduction

In many developed countries, a leading cause of death is cancer that can be reduced by the detection of tumor at an early stage. Existing imaging modalities for cancer diagnosis, X-ray, CT (computerized tomography) and PET (positron emission tomography) involve the risk of radiation exposure. While MRI (magnetic resonance imaging) provides 3D structural information, it is costly and image acquisition time is relative long. Recently, molecular imaging has emerged to find clinical applications [1, 2]. However, the majority of molecular imaging probes require injection of fluorescence dye or targeting probes into a subject that hinders rapid clinical translation due to regulation and safety issues.

On the other hand, thermal imaging has been used for several years to analyze medical conditions associated with changes in the body or skin temperature without exogenous probes [3]. Originally, thermography has been used for portable non-destructive industrial inspection to inaccessible areas as a fast and inexpensive inspection technique to detect hidden defect [4]. Recently, thermography technology is found useful in biomedicine such as detection of breast cancer or melanoma without fluorescent dyes [5–8]. Thermographic scans show temperature distribution that is influenced by function, physiology, and metabolism of the tissue of interest. The temperature of skin surface depends on its emissivity, underlying blood supply, and ambient conditions as well as the thermal condition of a lesion surrounded by normal tissue. However, this passive thermal imaging modality is prone to give false-positive and its sensitivity is limited as the temperature differences are typically less than 1.5 $°C$ [9–11].

Recently, dynamic thermal imaging has emerged to provide further thermal properties of the specimen by quantifying thermal conduction following external thermal stimulation. When the skin surface cools down by external cooling, the differences in thermophysical properties of the lesion was used for detecting melanoma skin cancer or burn injury [6–8]. In addition to the methods, external irradiation can provide enhanced thermal imaging of selective heating of blood relative to surrounding water-rich tissue using LED source utilizing the strong absorption in the blood at 530 nm [12]. Their method was sensitive to changes in the tumor versus healthy tissue that was the result of properties of blood flow and vasculature. On the other hand, the contrast in our approach was due to the properties of blood vessels, the response of these vessels to heating or cooling, and the thermal properties of the tissue. So far there has been no systematic study on the active use of thermal modulation for tumor detection.

To overcome these limitations, we hypothesize that active thermodynamic modulation – controlled heating or cooling via convection - provides enhanced contrast between lesion and neighboring normal tissue. Presumably, this is due to the differences of the thermal condition and of thermo-physiological properties of the diseased tissue that may give rise to distinct patterns in response to carefully controlled thermal modulation.

While thermography approach has a great advantage of no need of exogenous contrast agents and no ionizing radiation exposure for tumor detection, conventional passive thermography still lacks enough contrast to be used for early tumor detection. Here, we demonstrate an advanced approach to providing active thermodynamic contrast by exploiting physiological responses of tumor tissues in response to controlled thermal modulation. We obtained several parameters of active thermodynamic contrast with convection heating and cooling modulation as well as conduction modulation that distinguishes tumors at two different growth stages from surrounding tissue.

2. Materials and methods

2.1 Setup of active thermodynamic contrast imaging

The setup of the active thermodynamic imaging is shown (Fig. 1). The thermal imaging experiments were performed in a room at a constant temperature of 23 ± 0.5 $°C$. The uncooled microbolometer-based thermal camera (FLIR A325sc, FLIR Systems Inc., USA) contains 320 × 240 pixels with a spectral response range of 7.5 – 13.0 um and noise equivalent temperature difference of 0.05 $°C$. The camera was positioned at 0.2 m above from the animal and collected data at a frame rate of 15 Hz. The field of view was approximately 12 cm × 9 cm. The SL4-DsRed tumor–bearing mice were also imaged with a home-assembled fluorescence imager after acquiring series of thermal images. The detailed information of the fluorescence imager was described previously [13]. In brief, the fluorescence imager used green excitation light (535 nm, TouchBright LED, Live Cell Instrument, Seoul, South Korea), and collected emission signal through an emission filter (579/34 nm bandpass) with a scientific CCD camera. The imaging scope and the achromatic lens combination form a magnified image on the sensor of the CCD camera. Regarding the spatial resolution, the fluorescence system has a resolution of 44 μm with the field of view of 7.5 mm at a working distance of 5 mm. On the other hand, one pixel of our thermal camera corresponds to 375 μm. Thus, the resolution of fluorescence imaging was higher than that of thermal imager which had a limited number of pixels. The tumor imaging was performed two times when the tumor was at its early-stage (< 100 mm³) and when it was established stage (> 100 mm³) over the course of tumor growth. Note that tumors with volume over 100 mm³ were frequently used for tumor treatment research such as investigate drug effect [14]. The tumor volume was calculated as ½*(length × width × height).

 

Fig. 1 Experimental setup of active thermodynamic contrast imaging (A) Schematic diagram of experimental setup. (B) Photograph of the setup consists of a thermal camera and a fluorescence imager. The thermal camera has a standard f 10 mm infrared imaging lens.

Download Full Size | PDF

2.2 Development of a murine xenograft tumor model

All experimental animal protocols were approved by Institutional Animal Care and Use Committee (IACUC) of Gwangju Institute Science and Technology (GIST-2017-058). Six-week-old Balb/C immunodeficient female nude mice (DBL, Eumseong, South Korea) were used. Each animal was subcutaneously injected with 1 × 10⁶ SL4-DsRed cancer cells in 100 μL volume into the right flank. To detect early stage tumors that are often invisible or not easily palpable, we used a fluorescent cancer cell line to localize and confirm the presence of tumor. The tumor xenograft model was established as previously described [15].

In accordance with the principles of the 3Rs (replacement, refinement, and reduction) of animals in research (https://www.nc3rs.org.uk/), we used three animals per group with a longitudinal follow-up. For the histological study, we used three animals for obtaining early-stage tumor samples and four animals for obtaining established tumor samples.

2.3 In vivo active thermodynamic contrast imaging using convection modulation

The experimental procedure is illustrated in Fig. 2(A). First, the steady state temperature distribution in the tumor tissue and neighboring tissue was recorded using a thermal camera. Next external thermal modulation was applied while measuring temperature changes from the tissue surface. Passive and active thermal imaging of tumors was performed one week and two weeks after tumor cell implantation. Finally, the time series of thermal images were quantitatively assessed to extract the rate of temperature change, and thermal recovery time or the relaxation time constant. The thermal modulation could be conduction, convection, or radiation. In this study, we used convection modulation for enhanced thermal contrast. Warm air flow at 42 °C or cold air flow at 28 °C was blown to the flank including tumor area for 10 sec with the velocity of 1.8 m/s. The velocity was measured by an air velocity meter (HHF11A, OMEGA Engineering Inc., USA).

 

Fig. 2 Active thermodynamic contrast imaging experimental procedure. (A) After generation of a subcutaneous tumor (SL4-DsRed cancer cell) to the right flank, the thermal imaging of sequential temperature distribution throughout the convection modulation on the skin with subcutaneous tumors. With the same animal, the data collection was performed when the tumor was at an early-stage and in its established stage using the thermal camera. After data collection, the time series of images were processed with custom-written software. (B) Region of interest of tumor and neighboring tissues. (C) An exemplary diagram for data analysis.

Download Full Size | PDF

To compare thermal responses of neighboring and tumor tissue, we selected four regions of interests (ROIs) of neighboring tissue (N1 –N4) located approximately the diameter of the tumor from its center (Fig. 2(B)). As tumors had essentially round shape, the tumor diameter was approximated as 2R = √(a × b), where ‘a’ and ‘b’ refer to the length and width of the tumor from elliptical geometry, respectively. We chose one ROI for smaller (i.e. early-stage) tumor and three randomly selected ROIs for bigger (i.e. established) tumor. Each sampling ROI consists of nine pixels (i.e. 3 × 3 pixels covering 1.13 mm × 1.13 mm), and their average temperature was used for analysis (Fig. 2(C)).

The goal of active thermal modulation was to warm-up or cool-down the skin surface sufficient enough to enhance the transient response differences between the neighboring and tumor tissue. After the convectional modulation for 10 sec, the temperature of tissue was returned to its background state (i.e. recovery phase) for 40 sec towards the room temperature. During the recovery phase, thermal images of the tumor and neighboring tissue were captured at 15 frames per seconds. Using the same animals, we collected data for the tumors at their early-stage and at their established stage when the tumors were grown (Fig. 2). All data were collected with the camera software, FLIR ResearchIR Max (FLIR Systems, Inc., USA). After post-processing steps, the temperature distribution of the tumor tissue and neighboring tissue was plotted as a function of time throughout thermodynamic modulation as shown in Fig. 3 and Fig. 4. The raw image data were saved in Excel spreadsheet and were analyzed using OriginPro (OriginLab Corporation, Northampton, MA, USA) software for time series of temperature changes in ROIs and using MATLAB (MathWorks, Natick, MA, USA) for creating the representative thermal images by matching the temperature scales with appropriate color bars.

 

Fig. 3 Thermal contrast imaging using convection thermal modulator with early-stage tumors. (A) Photograph of SL4-DsRed tumor-bearing mouse model. WL: white light image, FL: fluorescence image. (B, D) and (C, E) are representative sequential thermal images, and temperature changes over time in tumor and neighboring areas for heating and cooling, respectively from one animal. (F, G) shows the average and standard deviation in from all the animals in tumor and neighboring areas. The vertical shaded columns in F and G indicates the duration of thermal modulation for 10 secs for heating and cooling, respectively.

Download Full Size | PDF

 

Fig. 4 Thermal contrast imaging using convection thermal modulator with established tumors. (A) Photograph of SL4-DsRed tumor-bearing mouse model. WL: white light image, FL: fluorescence image. (B, D) and (C, E) are representative sequential thermal images, and temperature changes over time in tumor and neighboring areas for heating and cooling, respectively. (F, G) shows the average and standard deviation in from all the animals in tumor and neighboring areas. The vertical shaded columns in F and G indicates the duration of thermal modulation for 10 secs for heating and cooling, respectively.

Download Full Size | PDF

2.4 Histology analysis of early-stage and established tumors

To provide tissue-level mechanistic insight of the differences between early-stage and established tumors in response to external heating or cooling, histological analysis was performed. For histological analysis, we prepared additional tissue samples of three early-stage and four established tumors. Subcutaneous tumor tissues were excised en bloc and fixed in 4% (w/v) paraformaldehyde for 24 hours. The tissues were dissected, embedded in paraffin, and stained with hematoxylin and eosin. To evaluate vessel distribution and density, immunohistochemistry using smooth muscle actin antibody (clone HHF35, code M0635, DakoCytomation, Glostrup, Denmark) was performed on the same tissue block. 3µm-thick tissue sections were submitted for staining using an automated immunostainer (Bond-maX DC2002, Leica Biosystems, Bannockburn, IL, USA). Each specimen was examined by an experienced pathologist using standard published criteria.

3. Results

When the mouse flank containing tumor is actively heated or cooled in a controlled manner, quantitative analysis of the dynamic responses of tumor or neighboring tissue during heating or cooling phase and recovery phases from thermographic imaging provides parameters that contrast tumor lesion from normal tissue. The tissue surface temperature patterns differed from those present in the steady state condition that is observed during conventional thermal imaging. The proposed active thermodynamic measurement consists of three phases: (1) the initial phase when the skin is exposed to nominal ambient conditions, (2) the heating or cooling phase during active thermal modulation, and (3) the thermal recovery phase. The results suggest that thermodynamic contrast between the tumor and neighboring tissue can be enhanced by the external thermal modulation. After post-processing, the time changes of temperatures from neighboring areas and the tumor area were plotted in Fig. 3 and Fig. 4. These results demonstrate that there are notable differences between the tumor and neighboring tissue with active thermal modulation with convective heating and cooling modulation. In addition, the fluorescence images were used to localize and confirm the presence of tumor that was fluorescently labeled as the early-stage tumors were often invisible or not easily palpable.

3.1 Active thermodynamic imaging using convection modulation on early-stage tumors

The tumor area could be invisible to the naked eyes except for fluorescence image at its early stage when the size of the tumor is small approximately one week after tumor cell inoculation (Fig. 3(A)). To examine the extent to which tumors can be discriminated based on the temperature difference with the neighboring tissue, we took white-light, fluorescence, and thermal images of the xenograft tumor. Thermal images showed a slight skin temperature reduction around the inoculation site before any visible or palpable signs of a tumor (Fig. 3(B), 3(E) at t = 0 sec). After applying heating or cooling air flow to the target region, we found a distinct thermal response from image analysis regarding the rate of temperature change and thermal recovery time during the heating and recovery phase, respectively (Fig. 3(C)-3(D) and 3(F)-3(G)) between tumor and neighboring tissue.

During heating phase, the temperature rapidly increased in tumor region compared to its neighboring regions: $\Delta \text{T}$ (T_peak – T_initial) = $2.20\pm 0.60{}^{\circ }\text{C}$ in tumor region and $\Delta \text{T}$ = $1.30\pm 0.10℃$ in neighboring regions. When the temperature reached the high peak, the temperature difference was the highest (Fig. 3(D)). Likewise, when the temperature reached the low peak, the magnitude of temperature difference was the highest during the cooling phase $\Delta \text{T}$ = -$1.30\pm 0.10℃$ vs. $\Delta \text{T}$ = -$1.01\pm 0.15℃$. From the sequential thermal images, the temperature difference between the tumor and neighboring regions was the highest at the end of thermal modulation, i.e. at the 20 sec in both convective heating and cooling (Fig. 3(D) and 3(G)). The representative data (Fig. 3(B)-3(E)) were from one animal. The results from other mice were comparable, and the average data (Fig. 3(F), 3(G)) were from all three animals. The results suggest that thermal contrast between the tumor tissue and neighboring tissue are enhanced by both heating and cooling convection modulation.

3.2 Active thermodynamic imaging using convection modulation on established tumors

As we used longitudinal study design, the corresponding data in Fig. 4 were obtained at later time point (two weeks post-implantation) from the same animals used in Fig. 3. As the tumor grows further, its core temperature tends to decrease compared to those of neighboring tissue (Fig. 4(B), 4(E) at t = 0 sec). The temperature profile inside the estimated tumor boundary was extracted and used to calculate the average tumor temperature. With active thermal modulation, the temperature difference between tumor and neighboring tissue was bigger than that of the early-stage tumor.

During heating phase, the temperature rapidly increased in tumor tissue compared to that of neighboring tissue, with $\Delta \text{T}$ _tumor of $3.57\pm 0.40℃$ at tumor tissue and $\Delta \text{T}$_normal of $1.60\pm 0.20℃$. When the temperature reached the peak, the temperature difference was the highest (Fig. 4(D)). Similarly, during convection cooling modulation, the temperature decrease was greater in tumor with $\Delta \text{T}$_tumor of -$1.80\pm 0.30℃$ compared to that of neighboring tissue $\Delta \text{T}$_normal of -$0.90\pm 0.09℃$. From the sequential thermal images, the temperature differences between tumor and neighboring tissue were the highest at 20 sec (Fig. 4(G)).

3.3 Quantification of active thermodynamic contrast parameters

When using heating or cooling thermal modulations, the rates of temperature change and thermal recovery time – the active thermodynamic contrast parameters - displayed different patterns between tumor and neighboring tissue. The simple exponential fit ($\text{T}\left(\text{t}\right)=\text{a}+b{e}^{-t/\tau }$) in OriginPro (OriginLab Corporation, Northampton, MA, USA) software was used for obtaining the thermal recovery time $\text{(τ})$. The temperature change has calculated the difference between T_Peak and T_Initial (ΔT).

With the early-stage tumors, we found that the rate of temperature change (0.22 ± 0.06 $℃$/sec) was higher and thermal recovery time (τ = 7.30 ± 0.59 sec) was shorter for the tumor tissue comparing to its neighboring tissue (0.13 ± 0.01 $℃$/sec and τ = 11.91 ± 2.22 sec) using heating modulation. On the other hand, with the cooling modulation, the rate of temperature change was negatively larger (−0.13 ± 0.01 $℃$/sec) and thermal recovery time was longer (τ = 9.88 ± 1.13 sec) at tumor tissue than neighboring tissue (−0.10 ± 0.02 $℃$/sec and τ = 6.57 ± 1.11 sec).

With established tumors, the rate of temperature change was bigger than that of smaller size tumor (0.36 ± 0.03 $℃$/sec) and neighboring tissue (0.16 ± 0.02 $℃$/sec) using heating modulation. Also, using cooling modulation, the value was negatively larger in tumor than neighboring tissue (−0.18 ± 0.03 $℃$/sec vs. −0.09 ± 0.01 $℃$/sec). The thermal recovery time for tumor was shorter (τ = 8.07 ± 0.57 sec) than neighboring tissue (τ = 9.64 ± 0.51 sec) when heating to the skin. However, with cooling modulation, the recovery time at tumor tissue was longer (τ = 12.67 ± 2.36 sec) than that of neighboring tissue (τ = 3.11 ± 0.18 sec). All these comparisons showed statistical significance (p < 0.05) and displayed consistent results between the early-stage and established tumors (Fig. 5(A)-5(D) vs. E-H) while we note the limitation that only three animals were used for obtaining these data.

 

Fig. 5 The rate of temperature change and the thermal recovery time (τ) between tumor and neighboring tissue for active thermal imaging at xenograft mice model (n = 3). (A-D) Early-stage tumor and its neighboring tissue. (E-H) Established tumor and its neighboring tissue. (A-B) The comparison graph between early-stage tumor and neighboring tissue using heating modulation. (A) The rate of temperature change values. (B) Thermal recovery time during the recovery period. (C-D) The comparison graph between tumor and neighboring tissue using cooling modulation. (C) The rate of temperature change. (D) Thermal recovery time during the recovery period. (E-F) The comparison graph between established tumor and neighboring tissue using heating modulation. (E) The rate of temperature change values. (F) Thermal recovery time during the recovery period. (G-H) The comparison graph between established tumor and neighboring tissue using cooling modulation. (G) The rate of temperature change. (H) Thermal recovery time during the recovery period. The error bars are standard deviations from all the ROI data (i.e. total 3 for early-stage tumors and 9 for established tumors while 12 for all the neighboring tissues). * P < 0.05, ** P < 0.01, *** P < 0.001.

Download Full Size | PDF

In short, the temperature change was faster in tumor region than those in neighboring tissue. Or the tumor’s temperature changes faster to active heating or cooling modulation. This tendency was consistent in the early-stage tumors and the established tumors while the latter showed greater rate of temperature changes.

3.4 Histological features of the early-stage and established tumors

The histological data with additional tumor samples (Fig. 6, Fig. 7), we analyzed the morphology and vessel density at peripheral and central lesions from tumors. The tumoral capillaries in early-stage tumors’ peripheral and central lesions were not much different regarding their morphology and vessel density (mean ± standard deviation: 10.4 ± 2.5 vs. 8.8 ± 2.2 vessels per field of view (FOV), p = 0.31, five FOVs) shown in Fig. 6. The blood vessels in established tumors show contrasting features: irregularly dilated staghorn-like blood vessels with a higher density in the periphery and near-completely collapsed slit-like vessels in the tumor center (12.8 ± 1.5 vs. 5.4 ± 1.3 vessels per FOV, p < 0.0001, eight FOVs) shown in Fig. 7.

 

Fig. 6 Histological features of the early-stage tumor. (A-C) Hematoxylin & Eosin (H&E) stain, (D-E) Immunohistochemical stain using smooth muscle actin antibody to highlight vascular structures. (magnifications A: × 80, B, C: × 200, D, E: × 400). Early-stage tumors (n = 3, size ranges 2.2 – 4.2 mm in diameter) show little signs of tumor necrosis (only one out of three tumors has localized necrosis). The tumor periphery (Box 1, B and D) shows relatively loose cell arrangement with tumoral capillaries highlighted with immunohistochemistry, while the central region (Box 2, C and E) shows compactly arranged tumor cells with slightly decreased capillary density with some collapsing signs in E.

Download Full Size | PDF

 

Fig. 7 Histological features of the established tumor. (A-C) Hematoxylin & Eosin (H&E) stain, (D-E) Immunohistochemical stain using smooth muscle actin antibody. (magnifications A: × 40, B, C: × 200, D, E: × 400). Established tumors (n = 4, size ranges 7.9 – 12.5 mm in diameter) show extensive tumor necrosis for all the tumor samples. The tumor periphery (Box 1, B and D) shows irregularly dilated staghorn-like blood vessels while the central region (Box 2, C and E) shows densely packed tumor cells with mostly collapsed vessels of low density that are barely recognized with immunohistochemistry.

Download Full Size | PDF

4. Discussion

This study suggests that thermography with active thermal modulation, termed as active thermodynamic contrast imaging (ATCI), allows us to detect early-stage tumor with further contrast compared to passive thermography that depends on rather small temperature differences between tumor and neighboring tissue. With controlled convection heating and cooling modulation, we demonstrated ATCI’s utility using a xenograft tumor model.

With active heating modulation, the rate of temperature change was higher and relaxation time during thermal recovery was shorter than those of neighboring tissue. Interestingly, with active cooling modulation, the relaxation time during thermal recovery for tumor tissue was longer. During the active heating or cooling phase, the change of temperature was consistently faster in tumor tissue. However, the recovery behaviors after heating or cooling were contrasting, that is, shorter recovery time from heating and longer relaxation time from cooling. These results imply that there may exist different thermophysiological mechanisms involved when the tumor tissue recovers from heated or cooled states to its original conditions in comparison with its neighboring tissue. The underlying assumption for the calculation of recovery time constant is based on exponential model such as relaxation time [16].

Interestingly, when the size of the tumor was bigger or more established, these thermal contrasts were increased (Fig. 4 vs. Fig. 3). Tumor blood vessels can react differently than normal tissue. When heated, normal blood vessels dilate and drain the remaining heat out of the tissue. This might be why the temperature of normal tissue changes less than that of tumor tissue in both heating and cooling. On the contrary, the abnormal angiogenic tumor blood vessels cannot react to the external heating or cooling as quickly as normal blood vessels react. It is known that the thermal conductivity of skin and tumor are also different: thermal conductivity of skin and tumor are 0.31 W/m°C and 0.55 W/m°C, respectively [17]. The bigger thermal conductivity would be responsible for the faster temperature changes. Another aspect of this phenomenon may be explained by the lack of proper peripheral nerves in tumor tissue as thermoregulation requires an internal feedback mechanism to sense the disturbance of local tissue temperature to restore its original state to maintain thermal homeostasis [18, 19].

From the histological analysis of tumor samples, we found some interesting pieces of evidence that may explain the observed differences in thermal responses. First, the current tumor tissue shows high cellularity compared to the neighboring tissue from all the samples suggesting the different thermodynamic properties. Tumoral areas consist of densely packed tumor cells, but the surrounding tissue is predominantly composed of loose connective tissue and occasional neural and vascular structures. Second, the newly formed tumoral vessels show different features in the early-stage and established tumors. Furthermore, all the established tumors contain extensive necrosis with a geographic pattern that could contribute the different thermal response in addition to the blood flow differences while only one relatively small area of necrosis was found in one of early-stage tumors (Appendix, Fig. 8).

During image acquisition, the movement artifact of the animal from breathing and heart beating was unavoidable that resulted in limited accuracy of temperature measurement. In the future, we plan to resolve this issue and improve the quality of data with reduced fluctuations. Some researchers showed tracking approach to compensate the motion artifact of the human skin surface with adhesive markers at the center of the lesion and neighboring tissue based on the template algorithm [20]. Thus, we plan to develop motion artifact correction by introducing an infrared fiducial marker. It would also be meaningful to compare the size of tumor based on the fluorescence-based imaging and thermal contrast-based imaging in further experiments.

Regarding the specificity of this approach for tumor detection, it should be noted that the proposed active thermodynamic imaging cannot provide diagnostic information whether the lesion found by ATCI is cancer or not as other pathologies such as inflammation process might also give similar thermal response. Also, whether ATCI can correctly delineate tumor margin requires further investigation. Thus, this approach can be better suitable for screening purpose rather than diagnostic purpose shortly until more detailed potential is worked out. However, given its non-invasive and label-free nature, this simple and low-cost methodology may find a good position for initial tumor screening in the clinical setting.

While we demonstrated the ATCI approach with a subcutaneous tumor model, it will be more relevant and powerful if this approach also works with orthotopic or metastatic tumors growing inside the gastrointestinal tract or intraperitoneal cavity via endoscopic modality. The optimal type of thermal modulation – convection, convection, or radiation – needs to be tested for the endoscopic version of ATCI, and further investigation is warranted for this emerging field of cancer screening.

In summary, this study provides a significant progress in several aspects. Our approach used thermal modulation with ‘convection mode’ for the purpose of tumor detection even at its early stage when the tumor is not easily palpable or visible. Also, we tested both active ‘heating’ and ‘cooling’ convection modes to provide a comprehensive understanding of the different thermal responses of tumor tissue. Furthermore, with histological analyses of both H&E and vascular staining of early-stage as well as established tumor samples, we provide further evidence of the distinctive thermal contrast with tumor size regarding vascular distribution as well as tumor necrosis.

Appendix

 

Fig. 8 Evidence of tumor necrosis from early-stage tumor (A), and established tumors (B-E) with Hematoxylin & Eosin (H&E) staining (magnification: × 100). Only one relatively small area of necrosis was found in the early-stage tumors (A) while extensive necrotic changes were observed from all the established tumors (B-E, images from all four tumors).

Download Full Size | PDF

Funding

This work was supported by Biomedical Integrated Technology Research Project, GIST-Caltech Research Collaboration Project and the GIST Research Institute (GRI) provided by GIST in 2017. Also, supported by research grant (NRF-2016R1A2B4015381) of the National Research Foundation (NRF) funded by the Korea government (MEST), the Brain Research Program (NRF-2017M3C7A1044964) of the NRF, the KBRI basic research program through Korea Brain Research Institute (17-BR-04) funded by the Ministry of Science, ICT, and Future Planning.

Acknowledgment

We thank Mrs. Jihye Yang, Dr. Su Woong Yoo and Mr. Soonjoo Hwang (from Department of Biomedical Science and Engineering, Gwangju Institute of Science and Technology) for preparation of SL4-DsRed cancer cell line and animal care. Also, we thank Mr. M. Mohsin Qureshi for help with figures. We greatly appreciate Dr. Munseob Lee from the Electronics and Telecommunications Research Institute (ETRI) for lending us the thermal camera.

References and links

1. G. Oh, Y. Park, S. W. Yoo, S. Hwang, A. V. D. Chin-Yu, Y.-M. Ryu, S.-Y. Kim, E.-J. Do, K. H. Kim, S. Kim, S. J. Myung, and E. Chung, “Clinically compatible flexible wide-field multi-color fluorescence endoscopy with a porcine colon model,” Biomed. Opt. Express 8(2), 764–775 (2017). [PubMed]  

2. M.B. Sturm and T.D. Wang,” Emerging optical methods for surveillance of Barrett's oesophagus,” Gut, gutjnl-2013–306706 (2015)

3. N. A. Diakides, “Infrared Imaging: An Emerging Technology in Medicine,” IEEE Eng. Med. Biol. Mag. 17(4), 17–18 (1998). [PubMed]  

4. X. P. Maldague, Nondestructive evaluation of materials by infrared thermography (Springer Science & Business Media, 2012)

5. E.-K. Ng, “A review of thermography as promising non-invasive detection modality for breast tumor,” Int. J. Therm. Sci. 48(5), 849–859 (2009).

6. C. Herman, “The role of dynamic infrared imaging in melanoma diagnosis,” Expert. Rev. Dermatol. 8(2), 177–184 (2013). [PubMed]  

7. C. Herman and M. P. Cetingul, “Quantitative visualization and detection of skin cancer using dynamic thermal imaging,” J. Vis. Exp. 51, e2679 (2011). [PubMed]  

8. A. Renkielska, M. Kaczmarek, A. Nowakowski, J. Grudziński, P. Czapiewski, A. Krajewski, and I. Grobelny, “Active dynamic infrared thermal imaging in burn depth evaluation,” J. Burn Care Res. 35(5), e294–e303 (2014). [PubMed]  

9. C. Stefanadis, C. Chrysohoou, D. B. Panagiotakos, E. Passalidou, V. Katsi, V. Polychronopoulos, and P. K. Toutouzas, “Temperature differences are associated with malignancy on lung lesions: a clinical study,” BMC Cancer 3(1), 1 (2003). [PubMed]  

10. J. Wang, K.-J. Chang, C.-Y. Chen, K.-L. Chien, Y.-S. Tsai, Y.-M. Wu, Y.-C. Teng, and T. T.-F. Shih, “Evaluation of the diagnostic performance of infrared imaging of the breast: a preliminary study,” Biomed. Eng. Online 9(1), 3 (2010). [PubMed]  

11. J. F. Head, F. Wang, C. A. Lipari, and R. L. Elliott, “The important role of infrared imaging in breast cancer,” IEEE Eng. Med. Biol. Mag. 19(3), 52–57 (2000). [PubMed]  

12. J. R. Case, M. A. Young, D. Dréau, and S. R. Trammell, “Noninvasive enhanced mid-IR imaging of breast cancer development in vivo,” J. Biomed. Opt. 20(11), 116003 (2015). [PubMed]  

13. G. Oh, S. W. Yoo, Y. Jung, Y.-M. Ryu, Y. Park, S.-Y. Kim, K. H. Kim, S. Kim, S.-J. Myung, and E. Chung, “Intravital imaging of mouse colonic adenoma using MMP-based molecular probes with multi-channel fluorescence endoscopy,” Biomed. Opt. Express 5(5), 1677–1689 (2014). [PubMed]  

14. G. D. Ayers, E. T. McKinley, P. Zhao, J. M. Fritz, R. E. Metry, B. C. Deal, K. M. Adlerz, R. J. Coffey, and H. C. Manning, “Volume of preclinical xenograft tumors is more accurately assessed by ultrasound imaging than manual caliper measurements,” J. Ultrasound Med. 29(6), 891–901 (2010). [PubMed]  

15. S. W. Yoo, H.-J. Park, G. Oh, S. Hwang, M. Yun, T. Wang, Y.-S. Seo, J.-J. Min, K. H. Kim, and E.-S. Kim, “Non-ablative Fractional Thulium Laser Irradiation Suppresses Early Tumor Growth,” Curr. Opt. Photon. 1(1), 51–59 (2017).

16. A. Renkielska, A. Nowakowski, M. Kaczmarek, and J. Ruminski, “Burn depths evaluation based on active dynamic IR thermal imaging-a preliminary study,” Burns 32(7), 867–875 (2006). [PubMed]  

17. G. M. Hahn, “Metabolic aspects of the role of hyperthermia in mammalian cell inactivation and their possible relevance to cancer treatment,” Cancer Res. 34(11), 3117–3123 (1974). [PubMed]  

18. S. Li, Y. Sun, and D. Gao, “Role of the nervous system in cancer metastasis,” Oncol. Lett. 5(4), 1101–1111 (2013). [PubMed]  

19. T. H. Benzinger, “Heat regulation: homeostasis of central temperature in man,” Physiol. Rev. 49(4), 671–759 (1969). [PubMed]  

20. T.-Y. Cheng and C. Herman, “Involuntary motion tracking for medical dynamic infrared thermography using a template-based algorithm,” Proc. SPIE 86692, 86692Q (2013).