Lasers in Medical Science

, Volume 21, Issue 4, pp 192–197

Raman spectroscopy of parathyroid tissue pathology

Authors

  • Kaustuv Das
    • Department of Breast and Endocrine SurgeryGloucestershire Royal Hospital
    • Biophotonics Research Group
  • Catherine Kendall
    • Biophotonics Research Group
  • Clare Fowler
    • Department of Breast and Endocrine SurgeryGloucestershire Royal Hospital
  • J. Christie-Brown
    • Dept. of PathologyGloucestershire Royal Hospital
Original Article

DOI: 10.1007/s10103-006-0397-7

Cite this article as:
Das, K., Stone, N., Kendall, C. et al. Lasers Med Sci (2006) 21: 192. doi:10.1007/s10103-006-0397-7

Abstract

Primary hyperparathyroidism (HPT) in 80% of patients is due to a solitary parathyroid adenoma, while in 20% multigland pathology exists, usually hyperplasia [Scott-Coombes, Surgery, 21(12):309–312, 2003]. Despite recent advances in minimally invasive parathyroidectomy, better preoperative localisation techniques and intraoperative parathyroid hormone (PTH) monitoring, a 4% failure rate [Grant CS, Thompson G, Farley D, Arch Surg, 140:47–479, 2005] persists making accurate differentiation between adenomas and hyperplasia of prime importance. We investigated the ability of Raman spectroscopy to accurately differentiate between parathyroid adenomas and hyperplasia. Raman spectra were measured at defined points on the parathyroid tissue sections using a bench-top microscopy system. Multivariate analysis of the spectra was carried out to construct a diagnostic algorithm correlating spectral results with the histopathological diagnosis. A total of 698 spectra were analysed. Principal-component (PCA)-fed linear discriminant analysis (LDA) used to construct a diagnostic algorithm. Detection sensitivity for parathyroid adenomas was 95% and hyperplasia was 93%. These preliminary results indicate that Raman spectroscopy is potentially an excellent tool to differentiate between parathyroid adenomas and hyperplasia.

Keywords

HyperparathyroidismRaman spectroscopyDiagnosisHyperplasiaAdenoma

Introduction

Within the general population, one in 500 women above age 40 and one in 2,000 men of all ages suffer from primary hyperparathyroidism (HPT). In 80 to 90%, this is due to a solitary parathyroid adenoma, while in 10 to 20% multigland pathology exists, either hyperplasia or adenomas or both [1].

The cornerstone of management of primary HPT is surgery; the classical bilateral four-gland exploration with or without intraoperative frozen section gradually evolving, especially where adenoma is concerned, into a more focussed minimally invasive approach. This has been possible with better preoperative localisation techniques and intraoperative parathyroid hormone (PTH) monitoring.

The current gold standard method to differentiate adenoma and hyperplasia is histopathology. Intraoperative frozen section analysis is still used by many surgeons to confirm the diagnosis, though the time taken to do this is a drawback. The role of the intraoperative frozen section is to confirm parathyroid tissue and an indication of the underlying pathology as well [2].

Pathologically, adenomas are oval, lobulated, encapsulated tumours [35]. On a cut section, they appear greyish brown with foci of haemorrhage and calcification and may occasionally undergo cystic change [6]. On microscopy, the encapsulated tumour is very cellular, with an identifiable rim of compressed normal parathyroid tissue visible in about two-thirds of cases. The adenoma is predominantly made up of chief cells though any of the usual cell types seen in a normal parathyroid gland can occur [7]. Figure 1 is a section of a parathyroid adenoma from one of the study samples. The section has been stained with haematoxylin and eosin (H&E). The section demonstrates chief cells in a diffuse pattern with a capsular rim, outside which is a thin margin of compressed normal parathyroid tissue. In most cases, the other parathyroid glands usually appear normal or even suppressed or atrophic and often the presence of a microscopically normal second gland is the best evidence that a given parathyroid lesion is an adenoma [8].
https://static-content.springer.com/image/art%3A10.1007%2Fs10103-006-0397-7/MediaObjects/10103_2006_397_Fig1_HTML.jpg
Fig. 1

H&E-stained section of adenoma with a rim of normalparathyroid

Hyperplastic glands are primary or occur secondary to impairment of renal function or chronic malabsorption and are usually chief cell hyperplasias. Classically, in primary hyperplasia, all glands are enlarged, sometimes exceeding 10 g and appear reddish-brown in colour [9]. However, this is not always the case, and sometimes only one gland appears enlarged and nodular, while the others look normal. This variant can obviously be confused grossly with an adenoma. Additionally, it is also possible for all four glands to appear normal in size during surgery but are actually hyperplastic on histological examination [10]. On microscopy, the predominant cell type is the chief cell either in a diffuse or nodular cellular arrangement. Figure 2 is an H&E section of a hyperplastic gland. Abundant chief cells in a diffuse arrangement are seen. Noteworthy is the absence of a compressed rim of normal gland.
https://static-content.springer.com/image/art%3A10.1007%2Fs10103-006-0397-7/MediaObjects/10103_2006_397_Fig2_HTML.jpg
Fig. 2

H&E-stained section, hyperplastic gland

In secondary chief cell hyperplasia, the glands may vary from normal to hugely enlarged, the former identifiable as hyperplastic only by virtue of their reddish colour and microscopic hypercellularity. On microscopy, again chief cells predominate, sometimes associated with nodular collections of oxyphil cells.

An obvious question from the above description would be: how could adenomas be accurately separated from hyperplasia? Pathologically, the morphologic distinction between chief cell hyperplasia and adenoma is not easy. Neither gross appearance nor cell type is reliable in differentiating between the two. The presence of a rim of normal parathyroid tissue around the tumour and the identification of at least one normal parathyroid gland are thought to represent the best evidence by which a diagnosis of adenoma can be made over that of chief cell hyperplasia [8]. But these are often difficult to assess, and pathological distinction in all cases continues to be a problem. With more and more glands being removed using a focussed approach, the potential for failure is greater. The failure rate in the best of hands is around 3–4% [11]. This may even rise as these methods are universally adopted. Black and Utley have suggested that adenoma and chief cell hyperplasia merely represent different morphologies of the same process [12], but the surgical management of adenomas and hyperplasia differs, and therefore differentiation is still of importance.

Accurate differentiation of a gland into a hyperplastic or adenomatous one, histologically or otherwise, is therefore of paramount importance to ensure continuing success in minimally invasive surgery.

There is, therefore, a need for a superior methodology to establish a definitive diagnosis. A possibility, yet to be evaluated, of distinguishing these differing pathologies is their biochemical signatures.

Raman spectroscopy is a potential optical diagnostic technique that measures the inelastic scatter of light. Change in the vibrational and rotational frequency of a molecule induced by an excitation source (laser light) produces a frequency shift, which is measured, analysed and is specific to the molecular constituents of the studied sample. The spectrum thus obtained is referred to as the optical biochemical signature of the sample, making Raman spectroscopy one of the truest forms of optical histochemistry.

Using Raman microspectroscopy, biological tissues have been analysed with varying successes.

Raman spectroscopy of biological tissue has largely been directed at differentiating malignant from benign tissue and also in the pathogenesis of disease. Developments in instrumentation have enabled identification of tissue spectra rapidly and with ease. Biochemical signatures of proteins, lipids and nucleic acids have been identified and used to differentiate tissue pathologies and construct biochemical maps. Cancers of the oesophagus, lung, bladder, breast and cervix have been investigated [1317]. Kendall et al. have elucidated the spectral characteristics of normal, neoplastic and Barrett’s oesophagus and demonstrated a 94% sensitivity and 93% specificity for model spectral discrimination [16]. Haka et al. [18] have studied ex vivo samples of normal, fibrocystic, fibroadenomatous and malignant breast tissue from 58 patients. Data were fit using a linear combination model in which nine basis spectra represented the morphological and chemical makeup of the studied breast tissue. Their resulting diagnostic algorithm achieved 94% sensitivity and 96% specificity for distinguishing cancer from other tissues. Similar reports on other biological tissues exist in literature. However, no group has to date analysed parathyroid tissue pathology, nor have benign pathologies been compared for spectral differences. It is, therefore, reasonable to suggest that Raman spectroscopy would be eminently suitable as a diagnostic interrogative tool to differentiate parathyroid adenoma from hyperplasia. In addition its rapid acquisition times would imply the possibility of real time tissue diagnosis using biochemical tissue signatures.

We evaluated Raman microspectroscopy as a means to this end.

Materials and methods

Patient selection

Ethical approval for the study was obtained from Gloucestershire Local Research Ethics Committee. All patients with HPT admitted for a surgical neck exploration were invited to participate in the study. Any patient unable or unwilling to give a written informed consent was excluded from the study. Consent was obtained in the morning of the surgery and duly noted in the patient records.

Sample collection

Parathyroid tissue was collected during surgery. The entire diseased gland was removed and sent to the histopathological department for routine processing. No form of storage media was used during transport. The samples were snap frozen in liquid nitrogen, mounted on cryostat chucks and cut to 25- to 30-μ thick sections. These were then placed on calcium fluoride slides and stored at −80°C for subsequent analysis. Calcium fluoride was used, as the spectrum of this material does not interfere with the spectral peaks of biological tissues. Part of the remaining tissue was used for fresh sample tissue analysis, while the rest underwent standard tissue processing. The use of frozen tissue was to duplicate as near enough as possible an in vivo approach. During standard tissue processing cut sections of the sample tissue were mounted on glass slides and stained haematoxylin and eosin (H&E) stain for pathological analysis. A single histopathologist confirmed the histology of each sample.

A total of 15 glands were sampled, nine of which were adenomatous and six hyperplastic.

Tissue spectral measurement

Near-infrared Raman spectroscopy was used for tissue spectral analysis. The spectroscope was a bench top Renishaw System 1000 spectrometer with a Leica DLM microscope and motorized XYZ stage, which allowed accurate pinpoint positioning of the tissue sample. A diode laser at 830 nm incident light was used as the excitation source. A ×80 0.75 NA Olympus NIRPlan ultra-long working distance (ULWD) lens focussed the excitation beam with a power of approximately 35 mW to a spot size of 3×10 μm at the sample.

Just before spectral analysis the tissue-mounted calcium fluoride slides were thawed. Homogenous areas from the adenoma samples and hyperplastic glands were identified for spectral analysis under low-power microscopy. After instrument calibration, Raman spectra were recorded from between 30 and 60 points from within these areas in each sample as random spot measurements at 50- to 100-μm intervals using the ×80 objective.

Spectral maps (i.e. spectra taken at regular intervals across the sample) were collected from samples where possible. All spectra were acquired for 30 s with two accumulations.

Spectral analysis

Multivariate analysis was used to develop an algorithmic model. From selected regions spectral peaks in the 400 to 2,000 cm−1 wave number range were analysed.

Spectra were loaded into the Matlab 6.1 statistical analysis programme and principal component scores (PCA) generated. These scores reflected the maximum points of variation in the data set and is a commonly used data reduction technique. The variation between the two pathological groups can be maximised using linear discriminant analysis (LDA). The PCA scores were used to develop a linear discriminant model, utilising PLS-toolbox (Eigenvector) and in-house developed tools.

Results

A total of 698 Raman spectra were recorded from these samples, 513 from the nine adenomas and 185 from the six hyperplastic glands. The averaged Raman spectra of adenoma and hyperplastic tissue are shown in Fig. 3.
https://static-content.springer.com/image/art%3A10.1007%2Fs10103-006-0397-7/MediaObjects/10103_2006_397_Fig3_HTML.gif
Fig. 3

Averaged point spectra from the two pathologies

Comparing the averaged spectra of the two pathologies, the following observations and inferences can be drawn:

On visual inspection, the spectra appear to be very similar. However, subtle differences are present. The most noticeable difference lies in the Amide III band at 1251 cm−1 for adenoma and 1257 cm−1 in hyperplasia. In addition to a shift to a higher wave number in hyperplasia, which indicates a change in protein secondary structure, the intensity is greater in the adenoma group particularly if compared with the adjacent lipid peak at 1298/1306 cm−1. Figure 4 compares the average point spectra of the two pathologies after background subtraction enabling a clearer visualisation of differences in peaks. Protein peak intensities at 542 cm−1, 1439/1441 cm−1 and 1654 cm−1 were greater in the adenoma group suggesting a more active proliferative process. Again, compared to hyperplasia the DNA peak for adenoma at 642 cm−1 and the nucleic acid peak at 1574 cm−1 were of a greater intensity. On the whole, there was also a shift in peak intensities to higher wave numbers for the hyperplastic glands.
https://static-content.springer.com/image/art%3A10.1007%2Fs10103-006-0397-7/MediaObjects/10103_2006_397_Fig4_HTML.gif
Fig. 4

Averaged point spectra from the two pathologies after background subtraction

Results of the LDA predictive model are shown in Table 1. The model demonstrated a sensitivity and specificity for parathyroid adenomas of 95 and 93%, respectively, while the sensitivity and specificity for parathyroid hyperplasia was 93 and 95%, respectively.
Table 1

Results of the predictive model

 

 

Pathological diagnosis

Predicted (Raman)

 

Adenoma

Hyperplasia

 

492

12

Hyper

21

173

A histogram of the linear discriminant function scores (Fig. 5) demonstrates a clear separation of the spectra.
https://static-content.springer.com/image/art%3A10.1007%2Fs10103-006-0397-7/MediaObjects/10103_2006_397_Fig5_HTML.gif
Fig. 5

Histogram of linear discriminant function scores

We cross-validated our results using a leave one spectrum out (LOSO) method, i.e. one spectrum from the dataset was held back and a multivariate model developed with the remaining spectra. The spectrum held back was then tested against this model. Results from this analysis can be seen in Table 2. Cross-validated results using the LOSO method demonstrated a sensitivity and specificity for parathyroid adenomas of 97 and 91%, respectively, and a sensitivity and specificity for parathyroid hyperplasia of 91 and 97%, respectively.
Table 2

Cross-validated results

 

 

Pathological diagnosis

Predicted (Raman)

 

Adenoma

Hyperplasia

 

500

13

Hyper.

16

169

Discussion

To our knowledge, this is the first such biochemical analysis of parathyroid pathology using Raman spectroscopy. This study has shown the ability of near-infrared Raman spectroscopy to differentiate between these two pathologies. The LDA model was able to demonstrate this with a high degree of sensitivity and specificity. The greatest drawback of this study is the small sample numbers, and further research aimed at resolving this is necessary. Raman mapping experiments also would aid in the diagnosis. On the whole, a clear separation of the pathologies was possible as evidenced by the histogram in Fig 5. A number of spectra [12] showed a degree of overlap, and this would possibly support the view stated by Black and Utley that adenoma and chief cell hyperplasia merely represent different morphologic manifestations of the same process [12]. However, further elucidation of the biochemical signature differences needs to be studied.

These preliminary results indicate that Raman spectroscopy could be a clinical tool to differentiate between parathyroid adenomas and hyperplasia. Should further sample analysis confirm this, Raman spectroscopy would then be an excellent adjunct to histopathological diagnosis.

Ultimately, we hope to prove the feasibility of intraoperative biochemical tissue analysis to differentiate between hyperplasia and adenoma using an in vivo Raman probe, which would support or replace intraoperative frozen sections for tissue pathology. With advances in development of probes and portable systems, this could soon become a reality. This would then translate into shorter operating times, potentially increased diagnostic certainty, and therefore decreased reoperation rates.

Also in general, the implications for real time tissue diagnosis using their biochemical signatures are enormous.

Acknowledgements

Dept. of Pathology Gloucestershire Royal Hospital for their help and support.

Dr Sare Paul Professor of Pathology (Retd) Madras Medical College, India.

Copyright information

© Springer-Verlag London Limited 2006