Introduction
Differentiation of normal skin from benign skin lesions and from cancerous tissue has been a major diagnostic challenge as the number of skin cancers continues to increase worldwide. Yet, early diagnosis of small lesions is important to provide a wider array of treatment options. Excisional versus topical treatments have become important issues especially for keratinocyte cancers. Depending on their size, shape, and rate of recurrence there may be different treatment options.1,2 While identification of basal cell carcinoma can be easily achieved based on visual and dermoscopic analysis, identification of lesion subtype can be further refined based on other imaging methods including ultrasound and vibrational optical coherence tomography, based on their size and shape.3,4
We recently developed a technique termed vibrational optical coherence tomography (VOCT) to noninvasively characterize different types of skin cancers.4–8 The results indicate that cancerous lesions are characterized by deposition of stiff fibrotic tissue, new blood vessels, and new cancer associated cells that are not present in the papillary and reticular dermis of normal skin.5–8 The finding of resonant frequency peaks consistent with the deposition of stiff fibrotic tissue is consistent with the reports that lysyl oxidase promotes crosslinking of stiff, highly collagenous tissue.9–12 The presence of fibrous tissue demonstrated by VOCT measurements may provide a method for early screening of both skin and other epithelial cell-derived cancers.
Beyond its role in cancer, organ fibrosis affects all major human tissues and can lead to excessive accumulation of extracellular matrix (ECM) components, predominantly collagens. Forty-five percent of all deaths in the developed world are believed to be associated with fibrosis.13 A clear relationship between chronic inflammation, fibrosis, and cancer dates back almost 200 years.14 The ability to noninvasively detect tissue and organ fibrosis is an important tool to allow early diagnosis of many fibrotic diseases.
Cancer associated fibroblasts (CAFs), but not normal fibroblasts, support metastatic lesions.15 CAFs and tumor cells are regulated by provisional matrix molecules, leading to continued stromal collagen cross-linking, and resulting fibrosis.10 CAFs are major components of the stroma surrounding carcinomas. New cells, as well as new blood vessels, are seen in cancerous skin lesions. For example, in basal cell carcinoma (BCC) vessels are seen with large diameters that are branching as well as superficial vessels (telangiectasia) consisting of short fine linear vessels with fine branches.16 While imaging methods can provide qualitative differences between normal blood vessels and cancer associated vasculature, VOCT can provide quantitative differences that can be used to differentiate benign from cancerous skin lesions.4–8 Besides measuring tissue component stiffness using VOCT, it is possible to create a “virtual biopsy” mapping the location of new cells and fibrotic tissue.6
Both cancer cells and skin cancerous tissue are stiffer than normal tissue.17–20 The increased stiffness of cancer associated cells and fibrous tissue found in skin carcinomas using VOCT has been reported previously.4–8 The technique applies audible sound and infrared light transversely to the skin surface providing an image and a mechanovibrational spectrum of the major skin components.4–8 Normal skin has major resonant frequency peaks at 50 Hz (stiffness about 1.2 MPa), 100 Hz (2.5 MPa), and 150 Hz (4.5 MPa), while cancerous lesions have new peaks at 80 Hz, 130 Hz, and 260 Hz.4–8 The 50 Hz (1.2 MPa) peak appears to reflect the behavior of epidermal cells and fibroblasts, while the 150 Hz (4.5 MPa) peak is like that seen in normal arteries and veins.21 The new peaks at 80 Hz (about 2.1 MPa) and 130 Hz (about 4.5 MPa) have been hypothesized to reflect the presence of new pre-cancerous and other cells (perhaps CAFs) that are associated with new blood friable blood vessels (130 Hz peak). The 260 Hz peak observed in all cancerous skin lesions is associated with the formation of fibrous tissue (about 16 MPa) and is much higher than the modulus of dermal collagen (about 2.5 MPa).4–8 By creating a vibrational spectrum of each skin lesion type, it is possible to develop a lesion fingerprint that can be used to noninvasively provide an early diagnosis of lesions as small as about 0.1 mm in diameter.
The purpose of this study was to compare the mechanovibrational peaks found in normal skin, actinic keratosis (AK), BCC, squamous cell carcinoma (SCC), and melanoma to identify differences that can be used to create a “fingerprint” of each type of skin lesion as an early screening test for skin cancer. These “fingerprints” may be used with machine learning algorithms to identify lesions that have metastatic potential before they grow in size. In addition, the ability to define the interface between normal skin and cancerous tissue noninvasively would facilitate complete lesion removal during surgery. The results reported in this study suggest that the mechanovibrational peaks observed by vibrational optical coherence tomography can provide a specific lesion “fingerprint” that can be used to noninvasively screen different skin carcinomas. The differences in the resonant frequency peak locations can potentially be used to locate the boundaries of cancerous tissues.
Methods
Subjects
Normal skin from 14 subjects (11 males and 3 females) was studied in vivo using VOCT after informed consent was obtained as reported previously.22 Resonant frequencies of the components of skin were measured in vivo.22 The subjects ranged in age from 21 to 71 years old. Previous VOCT study results suggest that the frequencies of mechanovibrational peaks are similar when measured either in vitro or in vivo.7
Lesions identified by dermoscopy in the Dermatology Clinic at Summit Health (Berkeley Heights, NJ) were biopsied and studied in vitro as previously reported.22 The biopsied lesions were studied blindly by VOCT in vitro without identification of the age and sex of the patient. All subjects signed consent forms prior to enrolling in the study. Pathologic diagnoses were compared to mechanovibrational peak heights observed by VOCT after correction for speaker vibrations and vibrations not in phase with the input sinusoidal sound wave.22
OCT images
OCT image collection was accomplished using a Lumedica Spectral Domain OQ 2.0 Labscope (Lumedica Inc., Durham, NC) as described previously and was correlated with the histological images seen in sections cut from tissue biopsies as reported elsewhere.4–8
Measurement of resonant frequency and elastic modulus
The OQ Labscope was modified by adding a 2-inch diameter speaker to vibrate the tissue in the VOCT studies and modified to collect and store single raw image data that was used to calculate sample displacements (amplitude information) from amplitude line data.4–8,21 The vibrations for each frequency were isolated to calculate the amplitude. These amplitudes were plotted against the frequency of the vibrations. Mechanovibrational spectra of all samples were normalized by dividing by the height of the largest peak.
The measured resonant frequencies were converted into elastic modulus values using a calibration equation (Equation 1).4–8 The peak frequency (the resonant frequency), fn is related to the elastic modulus, E in MPa and the sample thickness d in meters.
Soft Tissues E×d=0.0651×(fn2)+233.16
Modulus values listed in Table 1 were calculated using Equation (1).
Table 1Resonant frequencies and moduli of normal skin, AK, BCC, SCC, and melanoma were studied
Resonant Frequency
|
|---|
| Sample | No. of Measurements | 50 Hz | 80 Hz | 100 Hz | 130 Hz | 260 Hz |
|---|
| Normal Skin | 14 | 50 {0} | NA | 100 {0} | NA | NA |
| AK | 7 | 50 {0} | 75.71 {7.68} | 104.28 {5.34} | 127.14 {4.81} | NA |
| Nodular BCC | 27 | 52.77 {4.60} | 75.55 {5.06} | 102.22 {4.23} | 125.18 {5.09} | 262.59 {4.46} |
| Superficial BCC | 6 | 54 {5.47} | 80 {0} | 102 {4.47} | 126.67 {5.16} | 261.67 {4.08} |
| SCC | 46 | 49.56 {2.06} | 75.86 {4.97} | 104.78 {5.05} | 127.82 {4.17} | 262.82 {4.55} |
| Melanoma | 57 | 50 {0} | 76.31 {4.86} | 102.20 {4.18} | 128.24 {3.83} | 262.98 {4.61} |
| Normal Skin | 14 | 1.231 {0.140} | NA | 2.66 {0.32} | NA | NA |
| AK | 7 | 0.88 {0.160} | 1.60 {0.24} | 2.54 {0.81} | 4.52 {1.206} | NA |
| Nodular BCC | 27 | 1.467 {0.26} | 2.31 {0.41} | 3.20 {0.46} | 4.41 {0.78} | 16.51 {2.65} |
| Superficial BCC | 6 | 1.525 {0.74} | 2.27 {0.40} | 2.86 {1.207} | 4.46 {0.69} | 16.38 {2.82} |
| SCC | 46 | 1.304 {0.156} | 2.02 {0.28} | 3.15 {0.45} | 4.30 {0.47} | 15.68 {1.826} |
| Melanoma | 57 | 1.459 {0.162} | 2.26 {0.29} | 3.36 {0.41} | 4.81 {0.56} | 17.44 {1.938} |
Normal skin studies were conducted in vivo using a universal hand piece mount.22 Once VOCT studies were conducted, the biopsy samples were processed and reviewed by a trained Mohs dermatopathologist who conducted the pathological analysis.
Results
Results of experiments conducted on normal skin in vivo and excisions of AK, BCC, SCC, and melanoma in vitro indicate that there is a relationship between the organization of new cells at (80 Hz), new blood vessels (130 Hz), and fibrous tissue (260 Hz) in benign and cancerous lesions. Differences in the peak heights and frequencies can be used to differentiate among normal skin, AK, BCC, SCC, and melanoma based on the data shown in Figures 1–7 and Tables 2 and 3.
Table 2Peak height ratios and p-values for 50 Hz and 80 Hz mechanovibrational peaks for BCC, SCC and melanoma
50 Hz/80 Hz displacement ratios
|
|---|
| SCC | Melanoma |
|---|
| BCC | 0.038a | 0.37 |
| Melanoma | 0.031a | |
Table 3Peak height ratios and p-values for 80 Hz and 130 Hz mechanovibrational peaks for BCC, SCC, and melanoma
50 Hz/80 Hz displacement ratios
|
|---|
| SCC | Melanoma |
|---|
| BCC | 0.4 | 0.045a |
| Melanoma | 0.006a | |
Figure 1 shows a normalized mechanovibrational spectrum of normal skin as well as a cross-sectional OCT image of a typical area of normal skin. Data were normalized by dividing by the peaks due to the speaker and then by the highest peak in the spectrum. Note the peaks at about 50 Hz (cells), 100 Hz (dermal collagen), and 150 Hz (blood vessels) are typical of what is found in normal skin as reported previously.4–8
Figure 2 shows a normalized mechanovibrational spectrum of an AK (a), as well as an OCT image of the cross-section of the AK (b) where the measurements were made. The peaks at about 50 Hz (cells), 100 Hz (dermal collagen), and 150 Hz (blood vessels) are those found in normal skin. New peaks at 80 Hz and 130 Hz are found in AKs.
Figure 3 shows a normalized mechanovibrational spectrum of a typical nodular BCC (a), as well as an OCT image of the cross-section (b). Note the new peaks at about 80 Hz (cells) and 130 Hz (blood vessels) are similar to those found in AK and are supplemented by a new peak at 260 Hz that is not typically found in AK or normal skin.
Figure 1 illustrates that normal skin has prominent peaks at 50 Hz, 100 Hz, and 150 Hz, while AK has additional at peaks at 80 Hz and 130 Hz. Normal skin and AKs have different distributions of peak heights from BCCs, SCCs, and melanomas. AK has peaks at 80 Hz and 130 Hz, which distinguishes its profile from normal skin. It lacks a peak at 260 Hz, which differentiates AK from BCC, SCC, and melanoma (see Figs. 3–5).
Figure 4 shows a normalized mechanovibrational spectrum of a typical SCC (a), as well as an OCT image of the cross-section where the measurements were made (b). As shown in Tables 2 and 3, the peak height ratios for SCC are statistically different from those seen in BCC and melanoma.
Figure 5 shows a normalized mechanovibrational spectrum of a typical melanoma (a) and OCT images (b) of the cross-section of a melanoma where the measurements were made. Melanoma has different peak heights than are observed for BCC and SCC. As seen in the data shown in Tables 2 and 3, melanoma has peak height ratios that are statistically different when compared to BCC and SCC.
Figure 6 compares the normalized peak heights (normalized by dividing by the largest peak in a mechanovibrational spectrum of each sample) for normal skin, AK, BCC, SCC, and melanoma along with the standard deviations shown as error bars above the peaks. Peak locations, peak height differences, and peak height ratios can be used to differentiate differences in normal skin, AK, BCC, SCC, and melanoma as shown in Figures 6 and 7, as well as based on the data in Tables 2 and 3.
Figure 7 (a) shows the ratio of the 50 Hz and 80 Hz peak heights for BCC, SCC, and melanoma. The increased 50 Hz/80 Hz peak height ratio is significantly different for SCC compared to melanoma and BCC at a confidence level greater than 0.95, as shown in Table 2. Figure 7 (b) shows the relationship between the 130 Hz/80 Hz peak height ratios for BCC, SCC, and melanoma. The peak height ratio for melanoma is significantly different at a confidence level greater than 0.95 for BCC and SCC as indicated in Table 3.
Table 1 compares the stiffness of each peak shown in Figure 6, illustrating that the new cells at 80 Hz are stiffer than normal skin cells (50 Hz), and that fibrous tissue in cancerous lesions (260 Hz) is much stiffer than normal dermal collagen (100 Hz).
Discussion
A new technique termed VOCT, can be used to screen skin cancers noninvasively based on the morphology and stiffness of the lesions. The addition of new cells with increased stiffness (80 Hz), new blood vessels (130 Hz), and stiff fibrous tissue (260 Hz) that are present in all carcinomas studied can be used to differentiate benign from cancerous skin lesions.4–8,20–22 These results suggest that the product of fibrosis that results from cellular mutation is similar for different types of skin cancers, including cancers that have invasive and metastatic potential. While BBCs form primarily localized lesions, other cancers that are invasive can metastasize. The most difficult diagnostic task is to differentiate between BCC, SCC, and melanoma noninvasively. SCC appears to contain large amounts of normal cells and has a higher 50 Hz/80 Hz peak ratio, while melanoma has lower 130 Hz/80 Hz peak ratios, as indicated in Figure 7 and Tables 2 and 3. The relationship between the new cells (80 Hz), new blood vessels (130 Hz), and fibrous tissue (260 Hz) provides a means for noninvasively assessing the invasive and metastatic potential of a skin cancers based on the peak ratios of the 50 Hz/80 Hz and 130 Hz/80 Hz ratios.
The ability to predict the invasive potential of skin cancers noninvasively offers potential for early evaluation of lesions that may be difficult to diagnose. Using VOCT, it is possible to image and measure the stiffness of lesions as small as 0.1 mm (see OCT images in Figs. 1–5), which is smaller than can be seen by visual inspection or dermoscopy. VOCT results suggest that new cells (80 Hz peak), new blood vessels (130 Hz peak), and new fibrotic tissue (260 Hz peak) can be used to “fingerprint” cancerous tissues.
It has been reported that an epithelial–mesenchymal transition (EMT) results in a change of the cell phenotype of epidermal cells in epithelial derived cancers and could explain the new cellular peak at 80 Hz. During EMT, epithelial cells lose cell–cell adhesion attachments and reorganize their cytoskeleton. Cell-cell connections loosen, cells become motile, and become resistant to apoptosis.23 Transforming growth factor beta 1 (TGF-β1), tumor necrosis factor alpha, and hypoxia cooperate to trigger EMT both in cancer and fibrosis,23,24 converging in the induction of Snail activity through different mechanisms including the activation of transcription factor NF-kB.24–26 It is possible that the increase in the 80 Hz peak in AK, BCC, SCC, and melanoma is a consequence of EMT, which occurs during normal physiological processes such as wound healing and is not necessarily a marker for cancer since it is reversible.
CAFs become “activated” in tumors through interactions with cancer cells. They can promote tumor progression to malignancy and modify the stromal ECM by enhanced expression and activation of metalloproteinases.25 The cytokine TGF-β is one of the major prometastatic factors derived from CAFs.26 Whether the 80 Hz peak includes CAFs is unclear since AK contains a significant 80 Hz peak height. However, since AK is thought to be a pre-cancerous lesion the 80 Hz peak may be an important diagnostic sign of future formation of skin cancer.
CAFs secrete increased levels of TGF-β,26–31 which leads to the formation and deposition of dense fibrotic collagen fibers composed mainly of type I collagen (Col I). Col I is another prominent ECM protein with multiple contributions to tumorigenesis29 and has been reported to promote mammary tumor initiation.30 A prominent marker of all cancerous lesions studied is the deposition of fibrous collagen peak at 260 Hz with a stiffness of about 16 MPa, which is much higher than about 2.5 MPa reported for the collagen peak at 100 Hz in the dermis.4–8,20–22
Inflammatory markers are elaborated by endothelial cells that can sustain or exacerbate the inflammatory process. They recruit and activate leukocytes and have been postulated to lead to EMT transition that also contributes to fibrosis.31 While the 130 Hz peak is present in AK, BCC, SCC, and melanoma, it is highest in the invasive cancers, SCC, and melanoma, and may indicate a relationship between the appearance of new bloods vessels and invasive potential.
Based on the data in Figures 6 and 7 and Tables 2 and 3, it is possible to noninvasively differentiate among different types of skin cancer. The ability to classify lesions as small as about 0.1 mm in diameter should prove useful in assisting dermatologists in identifying potentially metastatic epithelial cancers and to promote nonsurgical methods to treat early stage benign and possibly malignant lesions. Ongoing studies using quantitative VOCT data in conjunction with machine learning may be useful to rapidly noninvasively identify different skin cancers.
Conclusions
In this study we have shown using VOCT that the morphology and stiffness of benign and cancerous lesions are different based on the addition of new cells with increased stiffness (80 Hz), new blood vessels (130 Hz), and fibrous tissue (260 Hz) present in carcinomas. By normalizing the peaks in the mechanovibrational spectra of the different skin lesions, differences in peak heights and peak ratios (50 Hz/80 Hz and 130 Hz/80 Hz) can be used to differentiate among AK, BCC, SCC, and melanoma and to noninvasively “fingerprint” these different skin lesions. Further studies involving machine learning will provide rapid noninvasive identification of skin lesion “fingerprints” for early screening of cancerous lesions.
It is possible that early diagnosis of small lesions can be enhanced and lead to more extensive use of topical treatments, which may be preferred for benign lesions, as wells as for BCC on the head, face and neck when excision may require extensive additional plastic surgery.
Abbreviations
- AK:
actinic keratosis
- BCC:
basal cell carcinoma
- CAF:
cancer associated fibroblasts
- COL(I):
collagen type I
- ECM:
extracellular matrix
- EMT:
epithelial mesenchyme transition
- SCC:
squamous cell carcinoma
- TGF-beta-1:
transforming growth factor beta-1
- VOCT:
vibrational optical coherence tomography
Declarations
Acknowledgement
The authors would like to thank Dr. Hari Nadiminti and coworkers at Summit Health, Berkeley Heights, NJ, for assistance in collecting the clinical data.
Ethical statement
All human studies were reviewed and approved by WCG IRB, approval code 1-1410030-1, 5 April 2021.
Data sharing statement
All data from this study can found at optovobronex.com.
Funding
None.
Conflict of interest
Mr. Tanmay Deshmuhk is an employee of OptoVibronex, LLC. Dr. Frederick H. Silver is a stockholder in OptoVibronex, LLC.
Authors’ contributions
Study concept and design (FHS and TD), performance of experiments (TD), analysis and interpretation of data (FHS), manuscript writing and review (FHS and TD).