1.

Introduction

Endoscopic ultrasound-guided fine needle aspiration (EUS-FNA) is increasingly used as a diagnostic modality to obtain a cytological diagnosis of suspect pancreatic masses. Surgery is currently the only treatment option for long-term survival in patients with pancreatic cancer. However, only a small minority of patients is eligible for surgery due to late detection, and it is often performed without pathological confirmation of the diagnosis.¹ Neoadjuvant therapy gains more interest to downsize the primary tumor and treat micrometastatic disease prior to resection.² Consequently, the need for a pathological confirmation is warranted before the start of chemotherapy.^3,4 Although EUS-FNA is a relatively safe procedure with a high specificity, it suffers from a high sampling error, resulting in a low-negative predictive value, and the need for an on-site cytopathologist.^5,6 Moreover, the accuracy of EUS-FNA drops considerably when (peritumoral) pancreatitis is present, which most frequently occurs.⁷

Several techniques have been investigated to enhance the diagnostic accuracy including contrast enhancement, confocal laser endomicroscopy, and optical coherence tomography.^8–12 However, these techniques involve administration of a contrast agent or extensive training to interpret the images. Fiber-optic spectroscopy is a relatively novel diagnostic modality that already showed its feasibility in a pilot study to characterize pancreatic tissue and to differentiate between normal and malignant tissue during surgery.¹³ Although these first in vivo results were encouraging, the size of the probe was not compatible with the small lumen of an FNA needle. No studies have yet been performed to assess the feasibility of fiber-optic spectroscopy during pancreatic EUS-FNA procedures.

Reflectance spectroscopy provides information about tissue absorption and scattering parameters. Light in the visible to near-infrared wavelength range is dominantly absorbed in the biological tissue by oxygenated and deoxygenated hemoglobin. Light is also absorbed by other chromophores, such as bilirubin and beta-carotene, depending on the tissue type. Single fiber (SF) reflectance spectroscopy uses only an SF to both emit the outgoing and capture the returning light. The advantage of using an SF is its ability to fit through an endoscopic biopsy needle, which, combined with its small sampling depth of approximately the fiber diameter,¹⁴ allows measurements at the exact biopsy location. Using a priori knowledge of the absorption spectra of the chromophores, a previously developed, described, and validated mathematical model can be used to extract and quantify physiological information, such as tissue oxygenation and blood volume fraction, from the obtained spectra.^14,15 In previous studies, our group has shown the potential of SF reflectance spectroscopy in detecting abnormal mediastinal lymph nodes during EUS-FNA procedures in patients with a clinical suspicion of lung cancer.¹⁶

In this study, we evaluated the feasibility of the incorporation of SF reflectance spectroscopy into pancreatic EUS-FNA procedures. The primary objectives were to assess the safety—assessed through adverse events—and ease of use—assessed through additional time taken—during the procedure. Moreover, we measured wavelength-dependent optical characteristics of pancreatic masses and assessed whether we could distinguish benign from malignant pancreatic tissue based on their optical properties.

2.

Material and Methods

2.2.

Measurement Setup

The experimental setup used in this study has been described in detail previously¹⁵ and is shown in Fig. 1. In short, the setup utilizes a single optical fiber connected to a quadfurcated optical fiber. Two arms are connected; one arm leading from a halogen light source HL-2000-FHSA (Ocean Optics, the Netherlands), and a second arm leading to a spectrophotometer SD-2000 (Ocean Optics, the Netherlands) to measure white light reflectance. A calibration procedure was performed to account for internal reflections, variability in lamp-specific output and in fiber-specific transmission properties.¹⁷

Fig. 1

Measurement setup. (a) The measurement device with laptop. The white arrow points at the connection point for the fiber. (b) The optical fiber, with a length of 3 m and an outer diameter of $700\mu \mathrm{m}$. (c) Incorporation of the optical fiber into 19-G puncture needle.

Sterilized single-use fibers (Leoni, Germany) had a core diameter of $300\mu \mathrm{m}$, an outer diameter of $700\mu \mathrm{m}$, an SMA905 connector, a distal polished fiber for wide-angle beam, a numerical aperture of 0.22, and a length of 3 m ($\pm 0.2\mathrm{m}$). The sampling depth depends on the optical properties but is typically similar to the core diameter of the fiber.¹⁴

3.

Results

3.1.

Patients

Sixteen patients undergoing a pancreatic EUS-guided FNA were included. In three patients no FNA, and thus no spectroscopic measurements, was performed, due to lack of visualization of the pancreatic mass. In the remaining thirteen patients, we were able to perform spectroscopic measurements. Of these, two patients were excluded due to a poor connection between the sterile measurement fiber and the quadfurcated optical fiber causing erroneous spectra and two patients were excluded due to lack of precise correlation with pathology (Fig. 2). In the latter two patients, the yield of the 19-G needle biopsies was inadequate and instead, good quality FNAs were acquired with the more flexible and smaller 25-G needle. One of the excluded patients was a patient with a pancreatic neuroendocrine tumor; the final puncture, obtained with a 25-G needle, was performed at the same location where the 19-G needle was positioned and spectroscopy measurements were taken. Because the cytology correlation might not be completely reliable, we described SF reflectance spectra of this patient separately. Characteristics of these patients are listed in Table 1.

Fig. 2

Inclusion scheme of patients.

Table 1

Patient characteristics. Patients 3, 6, and 12 were included, but no biopsy was taken so no SF reflectance spectroscopy measurements were performed.

Patient	Gender	Age	Site	Diameter tumor (mm)	Cytology
1	M	71	Corpus	30	Malignant: mucinous cystic tumor
2	M	74	Corpus/tail	80	Malignant: adenocarcinoma
4a	V	47	Head	NS	Neuroendocrine tumor
5	F	54	Head	NS	Malignant: metastasis adenocarcinoma ovary
7	F	70	Whole pancreas	50	Benign: fibrotic inflammation
8a	M	47	Head	25	Malignant: highly dysplastic cells
9a	M	63	Head and tail	NS	Benign: autoimmune pancreatitis
10	F	55	Tail	40	Malignant: adenocarcinoma
11	M	60	Corpus	33	Benign: autoimmune pancreatitis
13	F	60	Processus Uncinatus	NS	Malignant: adenocarcinoma
14	F	53	Processus Uncinatus	24	Malignant: adenocarcinoma
15a	M	69	Head/Processus Uncinatus	44	Benign: fibrosis
16	M	74	Head	20	Benign: fibrosis

Note: NS, not specified.

^aSpectroscopy measurements of these patients were not taken into account for the final analysis.

Cytological results could be correlated with spectroscopy measurements in the remaining nine patients, with cytological diagnosis of the spectroscopy-related FNA sample showing malignancy in six patients and benign disease in three patients. Two out of the three patients whose cytology showed a benign anomaly were at the end of their treatment plan suspect to suffer from pancreatic cancer. Patient #7 diagnosed with fibrotic inflammation, underwent pancreatic surgery based on clinical symptoms and preoperative images. The surgery was aborted due to hemodynamic instability. Patient #16, diagnosed with fibrosis, showed an increase in tumor volume and thickened lymph nodes on a CT-scan, performed one month after the EUS-FNA. Both patients died within four months after the EUS-FNA procedure. Although no final histopathological diagnosis could be made, this was most likely due to a pancreatic adenocarcinoma.

3.2.

Single Fiber Reflectance Spectra

No adverse events related to the spectroscopy measurements were observed. Additional endoscopy time related to the spectroscopic measurements was approximately five minutes.

Representative SF reflectance spectra from measurements taken at a benign and malignant spot (patients #16 and #2) are displayed in Fig. 3. In Table 2, the measured values of blood oxygen saturation, blood volume, and bilirubin concentration are displayed. Pancreatic tissue does not show much hemoglobin absorption features in the spectra, which are typically visible as “dips” in reflectance between 500 and 600 nm. The lack of strong absorption features in the spectra is reflected in the fitted values for the blood volume, which are low for both benign and malignant sites (1.1% and 2.2%, respectively; Fig. 4). The microvascular saturation is significantly higher at benign sites than malignant sites (55% versus 21%, $p=0.038$; Fig. 4). Bilirubin absorbs light mostly between 400 and 500 nm, with a peak around 450 nm. Benign sites were associated with significantly higher bilirubin content than malignant sites ($166\mu \mathrm{mol}/\mathrm{L}$ versus $17\mu \mathrm{mol}/\mathrm{L}$, $p=0.039$; Fig. 4). The amplitude and slope of the reduced scattering coefficient ($a_0$ and $a_1$, respectively) were not significantly different between benign and malignant tissue. Measurements of the patient who suffered from a pancreatic neuroendocrine tumor showed a blood level of $33.6\%\pm 1.6\%$ with a saturation of $86.2\%\pm 12.1\%$.

Fig. 3

Fitted spectra with corresponding cytology. (a) and (c) Spectra of a benign and malignant sample, respectively. The red line in the spectra indicates the actual measurements, and the black line shows the fit. The dotted line represents the expected spectra if no absorbers would be present in the sample. The residual—difference between measurement and fit—is shown below the spectra. (b) The corresponding cytology sample of (a): the lower left corner shows fibrosis (low magnification). (d) The corresponding cytology samples of (c): pancreatic ductal adenocarcinoma cells are shown in the middle (high magnification).

Table 2

Extracted means and standard deviations of blood volume, saturation, and bilirubin per patient.

Patient	Cytology result	Blood volume (%)		Saturation (%)		Bilirubin (μmol/L)
		Mean	SD	Mean	SD	Mean	SD
1	Malignant	0.7	0.3	29.4	45.9	2.5	1.8
2	Malignant	0.9	0.5	20.9	5.0	7.2	2.7
5	Malignant	1.4	0.4	0.0	7.8	11.8	3.1
7	Benign	0.3	0.1	48.3	30.3	39.2	18.1
10	Malignant	5.1	4.1	0.0	50.0	3.4	2.4
11	Benign	0.8	0.3	50.8	42.4	338.8	239.7
13	Malignant	3.1	1.9	49.6	4.2	26.6	0.5
14	Malignant	1.8	0.2	28.7	8.9	48.6	3.2
16	Benign	2.0	1.5	64.9	18.0	119.3	95.1

Note: SD, standard deviation.

Fig. 4

Difference in blood volume, saturation, and bilirubin concentration between cytological confirmed benign and malignant sites. Every individual patient is indicated by a dot. $P$-values are displayed above the scatterplots.

4.

Discussion

In this study, we showed that the incorporation of SF spectroscopic measurements during EUS-FNA procedures of the pancreas is feasible and safe. Because the fiber can be preloaded, measurements can be performed relatively easily and quickly. In our pilot study, the endoscopic procedure was prolonged with $\sim 5\min$ on average, which could be shortened after gaining more experience. No adverse events related to the spectroscopy measurements were observed, making it a safe procedure. Moreover, we demonstrated that optical properties, extracted from the SF reflectance spectra, appeared to be different between benign and malignant pancreatic aspirations. Therefore, incorporation of fiber-optic spectroscopy could potentially help to guide the FNA procedures and thereby reduce the sampling error and increase its negative predictive value without significantly affecting the complexity or time of the procedure.

Although EUS-FNA currently has a high specificity, it suffers from a low-negative predictive value implying that a negative result cannot be relied upon.²⁴ This is also confirmed by the current study: cytology results of three out of the nine patients showed a benign condition. Two out of these three patients (fibrotic inflammation and fibrosis), eventually turned out to have a malignancy by follow-up treatment or diagnosis and the FNA results thus represent sample errors. As more attention is given to neoadjuvant chemotherapy,² which requires a pathological diagnosis, it is highly important to reduce the sampling error and increase the negative predictive value of EUS-FNA procedures.

Pancreatic ductal adenocarcinoma is a hypovascular and a hypoxic tumor.^25–27 This is confirmed in this study, since both the benign and malignant sites showed very low levels of mean blood volume (1.1% and 2.1%, respectively). These results are supported by a study of Erkan et al.,²⁵ in which the microvascular density was determined: normal pancreas showed a five times higher microvascular density than fibrotic areas of chronic pancreatitis, and four times higher than pancreatic ductal adenocarcinoma. Interestingly, one of the patients in which we took measurements (but lacked good cytological correlation) was diagnosed with a pancreatic neuroendocrine tumor, which are known to be hypervascular.²⁸ That measurement indeed showed a far higher blood level of $33.6\%\pm 1.6\%$ with a saturation of $86.2\%\pm 12.1\%$.

The saturation of hemoglobin is a marker for the level of hypoxia.^29,30 We found significant lower mean saturation values in malignant areas compared to benign areas (21% and 55%, respectively), despite high overall arterial oxygen saturation values ($>97\%$). This is in agreement with results of Koong et al.,³¹ who intraoperatively measured pancreatic tissue oxygenation (using an Eppendorf ${\mathrm{pO}}_2$ histograph) and found significant tumor hypoxia, whereas normal tissue showed normal oxygenation. In another study, patients undergoing pancreatic surgery received pimonidazole—a hypoxia tracer that can be detected in tissue by immunohistochemistry—and showed that more hypoxia was present in the epithelial regions than in the stromal regions of the tumor (although there was a high intra- and intertumoral heterogeneity) and the absence of hypoxia in adjacent normal pancreatic tissue.³² Additionally, Lohse et al.³³ measured hypoxia in the stromal compartment of patient-derived pancreatic xenografts. They concluded that the levels of hypoxia in the tumor-associated stroma were generally lower when compared to the tumor compartment. These studies support our results. Although we did not measure normal tissue, we did—most likely—measure stromal areas of pancreatic adenocarcinoma in the two patients who suffered from fibrosis and inflammation (according to the cytology results obtained through EUS-FNA). These patients indeed showed a higher saturation and thus lower hypoxia levels than epithelial pancreatic adenocarcinoma regions. Furthermore, hypoxia is a known biomarker and potential therapeutic target in pancreatic cancer^34–36 and in vivo saturation measurements could provide interesting additional information to increase our understanding of the role of hypoxia before and during neoadjuvant treatment.

The levels of bilirubin found in this study are within the range that could be expected for this patient cohort. However, we cannot directly correlate the bilirubin concentration measured in pancreatic tissue to the bilirubin concentration measured in plasma, since we are only optically sampling a small amount of tissue with a low blood volume fraction. Moreover, we did not prospectively measure serum bilirubin levels prior to EUS-FNA. This would be a valuable addition as bilirubin levels can change quickly in cholestatic patients. It is interesting to speculate on the reason for our observation of lower bilirubin levels in malignant tissue compared to benign tissue. This may be a consequence of the (isolated) tumor microenvironment characterized by low blood volume, combined with the reduced availability of heme in rapidly proliferating tumor cells³⁷ and the possible degradation of bilirubin, known to be a strong antioxidant³⁸ in tumor cells under oxidative and (inflammatory) stress.

One other clinical study described in vivo spectroscopy measurements in human pancreatic tissue.¹³ It was concluded that a difference between normal and malignant tissue could be observed in collected reflectance at wavelengths around 470 and 650 nm, measured with a multifiber probe. However, the diameter of the multifiber probe is too large to fit through an endoscopic biopsy needle, making a translation to standard EUS-FNA procedures difficult.

In this study, reflectance spectra were analyzed using a fit algorithm [Eq. (1)] that incorporates the fact that the reflectance amplitude and the SF photon path length depend on the scattering properties of the tissue, with a dependence on not just the reduced scattering coefficient (${\mu }_s^{\prime }$) but also on the angular distribution of scattering (phase function) modeled through phase function parameter $\gamma$. We note that the use of Eq. (1) to fit the data no longer requires the use of the method described in a study by Kanick et al.³⁹ In that study, instead of explicitly expressing the $\gamma$ dependence of the collected SF reflectance, a set of empirical constants was used that minimized the error between the true and fitted SF path lengths. However, since the combined effects of ${\mu }_a$, ${\mu }_s^{\prime }$, and $\gamma$ on SF reflectance are now fully captured by Eq. (1), such an approach based on empirical constants is no longer necessary for accurate estimation of ${\mu }_a$. The fact that the different methods of fitting to Eq. (1) did not result in differences in the estimated ${\mu }_a$ of more than 6% indicates that this new method of fitting is robust.

A limitation of our study is the overall small sample size and small number of benign tissue measurements. However, our primary aim was to study the feasibility of incorporating SF reflectance spectroscopy measurements into the EUS-FNA procedure. Although it is encouraging that even with a small sample size differences in optical properties are observed, a larger (multicenter) study, including a wide variety of pancreatic lesions, is needed to fully exploit the potential of SF spectroscopy to guide the EUS-FNA sampling. In this study, we grouped pancreatitis with fibrous tumor parts (that were regarded as benign by the cytologist) together. More samples would enable us to also study the differences between these individual groups. The group of Hoffman et al., developed an orthotopic human tumor graft mouse model in which different fluorescent proteins were tested to image tumor progression.^40,41 This approach could be interesting as it allows visual distinction between pancreatic cancer cells and the surrounding stroma.^42,43 In addition, several clinical trials are currently investigating the safety and feasibility of intraoperative near-infrared fluorescence imaging using pancreatic cancer-specific contrast agents (NTR5673, NCT02736578).

The fiber used in this study fits through a 19-G needle, but is too thick for the more flexible 22-G needle, which is currently the preferred size at our gastroenterology department. Furthermore, we noticed that the SMA-connector of the sterile fiber was not always easily connected to the quadfurcated fiber, causing erroneous reflections in two patients. Future studies will, therefore, be performed with a smaller diameter fiber that fits through a 22-G needle, and with tighter tolerance requirements on the SMA-connector to secure a tight connection between fiber and measurement setup.

Real-time feedback is crucial to guide the gastroenterologist to epithelial tumor regions during the procedure, and, since our spectral fitting method is sufficiently fast ($<1\mathrm{s}$), this is feasible. Finally, to make the incorporation of SF reflectance spectroscopy into the FNA procedure even more simple, the small SF could potentially be integrated into the shaft of the biopsy needle, as already shown in a 15-G needle.⁴⁴ Integration of the optic fiber would allow direct visualization of the difference between benign and malignant tissue per patient, as we currently do not have data on intrapatient differences.

In conclusion, incorporation of SF reflectance spectroscopy measurements in EUS-FNA procedures of the pancreas is feasible and safe. Moreover, the optical properties differ between benign and malignant tissue sites, which could be used for real-time guidance of FNA sampling to reduce the sample error of these procedures and improve overall performance.