Differentiation between normal renal tissue and renal tumours using functional optical coherence tomography: a phase I in vivo human study

Authors


Kurdo Barwari, Meibergdreef 9, 1105 AZ Amsterdam, the Netherlands. e-mail: k.barwari@amc.uva.nl

Abstract

What's known on the subject? and What does the study add?

Optical Coherence Tomography (OCT) was developed in the early 1990s for ophthalmological application and is currently widely accepted in ophthalmology for retinal imaging purposes. In kidneys, the first experiments were performed on transplant kidneys to investigate the ability of OCT to assess ischaemic damage of kidneys. An ex vivo pilot study on the ability of OCT to differentiate normal renal tissue from malignant renal tissue, showed positive results and here we present the results of the first in vivo experiment.

The study shows for the first time that in vivo OCT is feasible and safe to perform in humans and that a significant different attenuation coefficient (as a quantitative measurement of the OCT images) is found between normal renal tissue and malignant renal tumours with a clear trend towards significance in the difference between benign and malignant renal tumours in a small pilot study. This suggests potential use of OCT in different clinical settings for diagnostic purposes in the course of renal tumours, and justifies further research.

OBJECTIVE

  • • To determine the ability of optical coherence tomography (OCT) in differentiating human renal tumours in an in-vivo setting by assessing differences in attenuation coefficient (µOCT; mm1) as a quantitative measurement.

METHODS

  • • Consecutive patients undergoing nephrectomy (partial/radical) or cryoablation for an enhancing solid renal tumour were included in our centre between October 2010 and May 2011.
  • • In vivo OCT images were obtained from renal tumour and normal parenchyma during surgery. Ex vivo OCT images of internal (subcapsular) tissue were obtained after longitudinal dissection of the extirpated specimen.
  • • Attenuation coefficients of the OCT images were determined off-line and compared between normal renal parenchyma and renal tumours (grouped per tissue type and per individual patient); and between OCT images recorded from tissue surface vs internal (subcapsular) tissue.

RESULTS

  • • In vivo OCT was performed in 16 cases (11 renal cell carcinoma, three benign tumours, one non-diagnostic biopsy and one not-accessible tumour).
  • • Median attenuation coefficient of normal renal parenchyma was 5.0 mm−1 vs 8.2 mm−1 for tumour tissue (P < 0.001) with normal parenchyma differing significantly from malignant tumour (9.2 mm−1, P < 0.001) and non-significantly from benign tumour (7.0 mm−1, P= 0.050). The attenuation coefficient of benign tumours did not differ significantly from that of malignant tumours (7.0 vs 9.2 mm−1, P= 0.139).
  • • Using patients as their own control, attenuation coefficients of normal renal parenchyma differed significantly from malignant tumour (P < 0.001) and non-significantly from benign tumour (P= 0.109).
  • • Assessed in 10 patients, there was no significant difference between attenuation coefficients of tumour surface and internal tumour (8.5 vs 9.7 mm−1 respectively, P= 0.260).

CONCLUSIONS

  • • In this first in vivo study on OCT for differentiation of renal tumours in humans the attenuation coefficients (as a quantitative assessment) differed significantly between normal renal parenchyma and malignant tumour.
  • • Tumour surface and internal tumour did not differ significantly, suggesting that a superficial OCT attenuation coefficient reliably assesses tissue composition inside the tumour.
  • • These results justify further research on OCT for various clinical applications in the diagnosis of renal tumours.
Abbreviations
OCT

optical coherence tomography

µOCT

attenuation coefficient

INTRODUCTION

The incidence of kidney and renal pelvis cancer is still rising, but a fast and reliable minimally invasive diagnostic method to establish a preoperative diagnosis of renal tumours is not readily available. In contrast to malignancies in most other organs, the high number of non-diagnostic results after renal tumour biopsy currently prevents general use of preoperative biopsies in the diagnostic evaluation of (small) renal tumours.

In a recent ex vivo pilot study we showed that optical coherence tomography (OCT) successfully distinguished normal renal parenchyma from malignant renal tumours, based on the optical properties extracted from the OCT images [1]. OCT is the optical equivalent of B-mode ultrasonography. Whereas ultrasonography is based on the intensity of time-delayed reflected sound pulses, OCT is based on the intensity of back reflected light. The intensity in both modalities is mapped to a spatial coordinate in the imaged specimen. OCT provides micrometre-scale resolution, cross-sectional images up to a depth of about 2 mm in renal tissue. Owing to this high resolution, OCT is usually visually correlated to structural information in histological images and therefore bears the potential to be the optical equivalent of normal biopsy (see Fig. 1) [2]. The imaging depth is limited by scattering of light by organelles and other cellular structures as the light penetrates the tissue, which hinders the return of reflections to the receiver. This attenuation of the OCT signal can be observed in images by decreased signal from greater depth and can be quantified by measuring the decay of signal intensity per unit depth using Lambert–Beer's law after careful calibration of the OCT system, resulting in an attenuation coefficient (µOCT, mm−1 ) [3]. Because malignant tissue displays more numerous, larger and more irregularly shaped nuclei with a higher refractive index and more active mitochondria, the attenuation of light is expected to be higher compared with normal and benign tissue. We hypothesize that by OCT signal analysis, measurable differences in µOCT between tissue types can be assessed [4,5].

Figure 1.

Example of an optical coherence tomography (OCT) image (A) of renal tissue with the corresponding histology as seen by microscopy (B). The comparable size of the blood vessels (black holes on OCT, white holes on microscopy) shows the similar order of resolution magnitude of both techniques.

This pilot study is, to our knowledge, the first to assess the ability of OCT to differentiate renal tumour tissue from normal renal parenchyma in an in vivo setting in humans. Furthermore, we aim to assess whether benign tumours can be differentiated from malignant tumours and whether superficial imaging by OCT is representative for the tissue located below the renal capsule, given the limited penetration depth of the technique.

METHODS

From October 2010 to May 2011, consecutive patients scheduled for nephrectomy (radical or partial) or laparoscopic cryoablation (for a solid enhancing renal mass suspect for renal cell carcinoma), were eligible for the study. Inclusion was based on informed consent approval by the patient and availability of the OCT device at the time of surgery. Institutional Review Board approval for this study was acquired.

The commercially available Santec Innervision 2000 OCT system (Santec Europe Ltd, Oxford, UK) used for this study acquired X images per second of 2 × 4 mm with 9-µm (depth) by 20 µm (lateral) resolution. The system was interfaced with a rotating sample arm probe that was developed in our institute for this study. The outer diameter of the probe was 2.3 mm and is therefore comparable with most common surgical trocars and endoscopes. The imaging direction of the probe is perpendicular to its axis of rotation (i.e. ‘sideways looking’). Before OCT imaging, the system response vs depth was carefully calibrated as described by Faber et al. [3].

Surgery was performed without deviation from the clinical protocol of the department. In the case of a radical nephrectomy, a small window was made in Gerota's fascia at the confluence of the tumour and normal renal tissue to provide access to the tumour surface.

The in vivo OCT probe was inserted in an optically transparent endo-ultrasonography cover under sterile conditions (see Fig. 2). After access to the tumour and normal renal parenchyma was provided, the surgeon (MPL) placed the tip of the OCT probe in contact with the normal renal parenchyma and the tumour. Five OCT images were recorded and stored, labelled with patient ID and type of tissue. In case of a laparoscopic approach, the covered OCT probe was introduced into the abdominal cavity through one of the trocars. By using the laparoscopic instruments, the tip of the probe was placed in contact with the kidney and tumour in a similar fashion to that used during open surgery and OCT imaging was performed as described above (Fig. 2). The overall image acquisition process took approximately 5 min, after which the normal surgical procedure continued.

Figure 2.

The in vivo optical coherence tomography probe covered in a sterile endo-ultrasonography cover used to obtain images from a renal tumour during laparoscopic surgery.

After extirpation, the specimen was transferred to the Department of Pathology for ex vivo measurements. The surface of the specimen was inked and cut longitudinally through the tumour to provide access to the tumoural internal tissue (see Fig. 3). Then, images were taken of the internal (subcapsular) areas of both tumour and normal parenchyma five times each by OCT, analogous to the in vivo procedure.

Figure 3.

Renal tumour longitudinally dissected to perform ex vivo optical coherence tomography imaging of the internal tumour tissue.

The attenuation coefficient (µOCT) was determined by one investigator (DMdB) who was blinded for pathology, by selecting a region of interest in the OCT image. In short, an average OCT signal vs depth was calculated (Fig. 4). By using the depth-dependent response of the OCT system from the calibration measurement, the µOCT of the region of interest was determined by the curve fitting of Beer's law.

Figure 4.

Optical coherence tomography (OCT) images of normal and malignant tumour tissue and the quantitative analysis procedure. First, a region of interest is selected in the OCT image indicated by the vertical blue lines. Second, the OCT signal vs depth within this region of interest is plotted in a graph. Finally, a descriptive mathematical model is fitted to this graph (transparent blue line) yielding the attenuation coefficient (µOCT) for both normal and tumour tissue. A steep slope will result in a high attenuation coefficient.

The mean attenuation coefficients from the five OCT images per tissue type per patient were stored in a PASW 18.0.2 database and combined with clinical parameters and demographic data from the corresponding patients. The standard pathological report as issued for clinical purposes was considered as the reference for tissue classification.

Statistical analysis was based on:

  • 1The difference of µOCT of normal tissue vs µOCT of tumours, including subcomparison of normal tissue vs malignant tumour (1a); normal tissue vs benign tumour (1b); and malignant tumour vs benign tumour (1c), all using a Mann–Whitney U test because of the non-normal distribution.
  • 2The difference of µOCT of normal tissue vs the µOCT of tumour per patient (e.g. using patients as their own controls), subdivided into normal tissue vs malignant tumour (2a) and normal tissue vs benign tumour (2b), using a Wilcoxon signed-rank test because of non-normally distributed paired measurements.
  • 3The difference of µOCT obtained from tumour surface vs internal tumour obtained after extirpation, using a Wilcoxon signed-rank test.

For all tests the differences were considered significant if the two-sided P value was <0.05.

RESULTS

From all cases that underwent radical nephrectomy or nephron-sparing surgery (partial nephrectomy or laparoscopic cryoablation) in our department in the study period, 16 cases were included and underwent peroperative in vivo OCT imaging. Overall, three tumours were benign (one oncocytoma, one leiomyoma, one benign cyst) and in one case no definitive pathological diagnosis of the tumour could be made (non-diagnostic biopsy result during laparoscopic cryoablation). In one case, the tumour could not be imaged during laparoscopy because of dense perirenal fat. Further details of the tumours and patients are shown in Table 1. For each OCT image the µOCT is determined as shown in Fig. 4.

Table 1.  Demographic data of patients and pathology of the 16 tumours
CharacteristicValues
Mean age (years)62.6 (30–81)
Mean tumour size (cm)4.3 (1.7–11.0)
Male : female11:5
Pathology 
 Clear cell renal cell carcinoma9
 Papillary renal cell carcinoma2
 Chromophobe1
 Benign3
 Non-diagnostic (biopsy during laparoscopic cryoablation)1

ANALYSIS 1

In vivo comparison of normal renal parenchyma vs tumour (per group). The tumour with a non-diagnostic biopsy result was excluded from this analysis, resulting in inclusion of the µOCT of 16 normal, three benign and 11 malignant OCT images.

The median µOCT of normal renal parenchyma was lower (5.0 mm−1) compared with tumour tissue (8.2 mm−1), P < 0.001. Among the three groups: (1a) median µOCT of normal tissue (5.0 mm−1) was lower than that of malignant tumour (9.2 mm−1), P < 0.001; (1b) the median µOCT of normal tissue (5.0 mm−1) was lower than that of benign tumour (7.0 mm−1), P= 0.050; and (1c) the median µOCT of benign tumour (7.0 mm−1) was lower than that of malignant tumour (9.2 mm−1), P= 0.139. The last two comparisons failed to reach statistical significance. The results are depicted in Fig. 5.

Figure 5.

Boxplot showing the median attenuation coefficient (µOCT) and interquartile ranges of in vivo acquired optical coherence tomography (OCT) images of the different tissue types. Normal µOCT= 5.0 mm−1 (4.3–5.4), Benign µOCT= 7.0 mm−1 (6.7–NA), Malignant µOCT= 9.2 mm−1 (7.4–9.9). The circle displays an outlier (patient 16, chromophobe renal cell carcinoma).

ANALYSIS 2

In vivo comparison of normal renal parenchyma vs tumour (per patient). Two cases were excluded from this analysis (non-diagnostic biopsy and the case without a tumour measurement), resulting in 14 patients with a µOCT of both normal tissue and the tumour. The results are depicted in Fig. 6. (2a) The median µOCT of normal renal parenchyma (5.0 mm−1) was lower than that of malignant tumour (9.2 mm−1), P < 0.001; and (2b) the median µOCT of normal renal parenchyma (5.4 mm−1) was lower than that of benign tumour (7.0 mm−1), P= 0.109.

Figure 6.

Attenuation-coefficients of in vivo acquired optical coherence tomography images of normal and tumour tissue displayed per individual patient. Patients 4, 10 and 11 had a benign tumour (oncocytoma, benign cystic lesion and leiomyoma, respectively). Patient 8 had a non-diagnostic biopsy result. In patient 12 the tumour was not accessible and patient 16 had a chromophobe renal cell carcinoma.

ANALYSIS 3

Comparison of superficial and internal recorded OCT images. Three cryoablation cases were excluded (no OCT imaging was possible because no tumour was extirpated) from this analysis. Three other cases were excluded because pathology could not be processed immediately after extirpation. In the remaining ten patients, no significant differences were observed between superficial and internal µOCT for both normal tissue (5.0 and 5.8 mm−1, respectively, P= 0.169) and tumour tissue (8.5 and 9.0 mm−1, respectively, P= 0.260). The results are depicted in Fig. 7.

Figure 7.

Boxplot showing the median attenuation coefficient (µOCT) and interquartile ranges of superficially and internal (subcapsular) acquired optical coherence tomography (OCT) images for normal renal parenchyma and tumours. Normal Superficial µOCT= 5.0 mm−1 (4.2–5.4), Normal Internal µOCT= 5.8 mm−1 (3.8–7.8), Tumour Superficial µOCT= 8.5 mm−1 (7.4–9.5), Tumour Internal µOCT= 9.0 mm−1 (7.6–14.4).The circle displays an outlier.

DISCUSSION

In this study we report the first results of OCT performed in renal tumours in vivo. Using quantitative assessment of the OCT images a significant difference between normal renal parenchyma and malignant renal tissue was established, proving the ability of OCT to differentiate in vivo normal and malignant renal tissue in real-time without violation of the tissue.

To overcome the obstacles in the diagnostic process of renal tumours mentioned in the Introduction, optical techniques are likely to be of additional value because they are light-based, and therefore harmless to human tissue – providing real-time information, and they are suitable for miniaturization, enabling integration with existing clinical procedures and instruments. OCT [2,6] is of specific interest because it provides a ‘functional optical biopsy’: cross-sectional images that can be correlated to histopathology in combination with quantification of optical properties that can be related to tissue physiology and cellular organization. Publications on the use of OCT in kidneys are scarce. In addition to ex vivo studies showing that renal microstructures can be visualized by OCT to assess ischaemic damage as an indicator of transplant kidney viability [7,8], we showed the ability of OCT to differentiate normal and malignant renal tissue by quantitative OCT assessment in an ex vivo pilot study [1].

Shortly thereafter, Linehan et al. [9] described microstructural differences of several renal tumours in ex vivo acquired OCT images, which allowed differentiation of several tumour types (e.g. angiomyolipoma and transitional cell carcinoma). Unfortunately, clear cell renal cell carcinoma and other renal cell carcinoma subtypes showed a heterogeneous appearance in the images, which precludes distinction of renal cell carcinoma from normal renal parenchyma based on the images alone. In a recent ex vivo multi-observer qualitative study Lee et al. [10] showed that OCT combined with confocal microscopy could differentiate normal from neoplastic renal tissue with a high sensitivity and specificity. Furthermore, they observed a marked decrease of imaging depth in tumour tissue caused by a higher degree of scattering, which is fully compatible with our hypothesis.

Translation of these promising ex vivo results to clinical measurements has never been proven and is not straightforward because of the variety of factors involved [11]. In more detail, the availability of suitable OCT probes is limited. We therefore developed our own prototype imaging probe for this study. The present probe configuration allowed two-dimensional imaging with a small field of view, which compromises image-based interpretation and comparison between tissue sites but did not hamper the ability to extract µOCT.

Moreover, most ex vivo results were obtained in the centre of the tumour (specimen) whereas in an in vivo setting only superficial measurements can be obtained. We therefore assessed whether superficial OCT imaging of tissue is representative for tissue that lies deeper. Comparison of µOCT obtained from the tissue surface with internal (subcapsular) areas after longitudinal dissection of the extirpated specimen showed no significant differences. This representativeness of superficial assessment using optical diagnostic tools with a limited penetration depth is also suggested by Bensalah et al. [12] using optical reflectance spectroscopy. Interestingly, the spread in µOCT values is higher for internal tissues, which we attribute to inhomogeneity of the central tumour.

The current in vivo study shows a statistically significant difference between values of µOCT for normal and malignant renal tissue, albeit still in a modest sample population. The small number of benign tumours precluded conclusions on the ability of OCT to differentiate benign from malignant tumours and normal tissue. Since the percentage (19%, three out of 16 extirpated tumours) is compatible with reports in the literature [13,14], this limitation is hard to avoid in this primary phase of clinical research. Moreover, the biology of the benign tumours varied widely, which makes interpretation of the determined µOCT challenging for these cases.

In spite of this limitation the current preliminary data on the ability of OCT to differentiate between malignant and normal renal tissue as well as the recently showed synergistic effect between OCT and Raman spectroscopy reinforce the investigation of quantitative OCT measurements in renal tumours. The ultimate goal should be to identify discriminatory thresholds in the attenuation coefficients between malignant and benign tumours.

If the diagnostic value of OCT is confirmed, the thin currently available OCT probes can be integrated with biopsy needles, enabling the incorporation of OCT into the diagnostic algorithm of renal tumours, and complementing or substituting percutaneous or operative biopsies. Furthermore, OCT could be used in the assessment of renal parenchyma margins after partial nephrectomy preventing frozen section analysis.

In conclusion, OCT provides high-resolution non-invasive cross-sectional images suitable for quantitative analysis. This phase I study showed that OCT can be safely employed in humans. A significant difference between attenuation coefficients of normal renal parenchyma and malignant renal tissue was found, proving the ability of OCT to distinguish malignant and normal renal tissue. Expansion of the population and validation of the results are needed to assess OCT as a clinically valuable new diagnostic tool.

ACKNOWLEDGEMENTS

This project was supported by the Cure for Cancer Foundation.

CONFLICT OF INTEREST

None declared.

Ancillary