ORIGINAL STUDY

Hyperspectral Imaging of Melanocytic Lesions Sudeep Gaudi, MD,* Rebecca Meyer, MD,† Jayshree Ranka, MS, MBA,‡ James C. Granahan, PhD,‡ Steven A. Israel, PhD,‡ Theodore R. Yachik, PhD,‡ and Drazen M. Jukic, MD, PhD§

Abstract: Hyperspectral imaging (HSI) allows the identification of objects through the analysis of their unique spectral signatures. Although first developed many years ago for use in terrestrial remote sensing, this technology has more recently been studied for application in the medical field. With preliminary data favoring a role for HSI in distinguishing normal and lesional skin tissues, we sought to investigate the potential use of HSI as a diagnostic aid in the classification of atypical Spitzoid neoplasms, a group of lesions that often leave dermatopathologists bewildered. One hundred and two hematoxylin and eosin-stained tissue samples were divided into 1 of 4 diagnostic categories (Spitz nevus, Spitz nevus with unusual features, atypical Spitzoid neoplasm, and Spitzoid malignant melanoma) and 1 of 2 control groups (benign melanocytic nevus and malignant melanoma). A region of interest was selected from the dermal component of each sample, thereby maximizing the examination of melanocytes. Tissue samples were examined at ·400 magnification using a spectroscopy system interfaced with a light microscope. The absorbance patterns of wavelengths from 385 to 880 nm were measured and then analyzed within and among groups. All tissue groups demonstrated 3 common absorbance spectra at 496, 533, and 838 nm. Each sample group contained at least one absorption point that was unique to that group. The Spitzoid malignant melanoma category had the highest number of total and unique absorption points for any sample group. The data were then clustered into 12 representative spectral classes. Although each of the sample groups contained all 12 spectral vectors, they did so in differing proportions. These preliminary results reveal differences in the spectral signatures of the Spitzoid lesions examined in this study. Further investigation into a role for HSI in classifying atypical Spitzoid neoplasms is encouraged. Key Words: hyperspectral imaging, melanocytic neoplasms, Spitz, atypical Spitzoid neoplasms (Am J Dermatopathol 2014;36:131–136)

INTRODUCTION Histopathologic examination of suspicious pigmented lesions remains the “gold standard” for the diagnosis of From the *Department of Pathology and Cell Biology, University of South Florida Morsani College of Medicine, Tampa, FL; †Department of Pediatrics, Children’s National Medical Center, Washington, DC; ‡Science Applications International Corporation, Surveillance and Reconnaissance Business Unit, Arlington, VA; and §Dermatopathology Service, James A. Haley Veterans’ Hospital, Tampa, FL. The authors declare no funding or conflicts of interest. Reprints: Sudeep Gaudi, MD, Department of Pathology and Cell Biology, University of South Florida Morsani College of Medicine, MDC 11, 12901 Bruce B Downs Boulevard, Tampa, FL 33612 (e-mail: sgaudi@ health.usf.edu). © 2013 Lippincott Williams & Wilkins

Am J Dermatopathol  Volume 36, Number 2, February 2014

cutaneous melanoma. Numerous architectural and cytologic attributes speak in favor of its diagnosis, and these features are generally sufficient for distinguishing malignant melanoma from benign melanocytic nevi. However, ambiguous (also known as “borderline”) cases do exist where inadequate criteria have led to significant diagnostic discordance among dermatopathologists.1,2 This deficiency in dermatopathology is perhaps most evident in the area of Spitz nevi, particularly in reliably classifying atypical Spitzoid neoplasms—lesions with features (Table 1) that reside in the esoteric realm between those of classic Spitz nevi and Spitzoid melanoma.3 Such ambiguous lesions are perhaps best referred to as “melanocytic neoplasms of uncertain malignant potential.” With this limitation in mind, researchers have been exploring adjunctive techniques to aid in distinguishing Spitzoid neoplasms. For instance, the analysis of DNA copy number changes by both comparative genomic hybridization and fluorescence in situ hybridization in Spitz nevi has consistently shown gains of chromosome 11p, an aberration not significantly reported in malignant melanoma.4–7 Although these observations are certainly intriguing and may eventually translate into routine clinical use, these techniques are not readily used in clinical laboratories due to issues with sensitivity, prolonged turnaround time, and low availability. Thus, dermatopathologists seek additional methods of aiding diagnosis. One such technique is hyperspectral imaging (HSI), a technology first developed for use in terrestrial remote sensing.8 HSI allows the identification of objects through the analysis of their unique spectral signatures—that is, the unique interplay between incident electromagnetic waves and the inherent physical properties of the object with which these waves interact. The National Aeronautics and Space Administration has used HSI for the detection and characterization of mineral deposits and air and ground pollution for many years. More recently, the literature has become replete with studies analyzing the potential application of this technology across numerous industries. For example, HSI has accurately distinguished viable from nonviable grains, potentially reducing pre-germination losses in agriculture9; automated nematode detection in cod fillets, conceivably improving quality control measures in commercial fishing10; and assessed the distribution of active pharmaceutical ingredients in tablets, theoretically ensuring effective delivery of the therapeutic agent.11 The potential use of HSI in the field of medicine is vast. Examples include early detection of foot ulceration in diabetics,12–14 identification of peripheral artery disease and assessment of its severity,15 detection of dental caries,16 and diagnosis of ocular diseases.17–21 Preliminary studies have endorsed use in various medical procedures, and results are encouraging. For www.amjdermatopathology.com |

131

Am J Dermatopathol  Volume 36, Number 2, February 2014

Gaudi et al

TABLE 1. Diagnostic Criteria Parameter

Classic Spitz Nevus

Atypical Spitzoid Neoplasm

Spitzoid Melanoma

Size Symmetry Lateral borders Irregular nesting Ulceration Deep extension Expansile nodule Maturation/zonation Deep border Cellularity Deep mitoses

Usually , 1 cm Symmetrical Sharply demarcated Uncommon Uncommon Uncommon Uncommon Common Infiltrating pattern Variable Rare or absent

Often . 1 cm Often asymmetrical Often poorly defined Common Variable Common Common Uncommon Rounded, pushing Increased in the dermis Common

Usually . 1 cm Asymmetrical Often poorly defined Frequent Common Variable Frequent Uncommon Irregular Increased Frequent

The criteria employed in diagnosing classic Spitz nevus, atypical Spitzoid neoplasm, and Spitzoid melanoma are listed here. Spitz nevus with unusual features (not listed here) was characterized by junctional predominance and somewhat greater cytologic atypia and/or pagetoid spread than typically allowed for a diagnosis of classic Spitz nevus.

example, assessment of renal oxygenation during partial nephrectomy may help minimize intraoperative ischemic injury,22,23 identification of tissue biomarkers may facilitate residual tumor recognition during surgery,24 and visualization of biliary system during cholecystectomy may minimize bile duct injury.25 Most relevant to the discussion at hand is preliminary data in the area of cancer diagnostics. HSI has shown potential use in the distinction between in vivo26,27 and in vitro28,29 benign and malignant tissues. The interest in this burgeoning technology has also extended to dermatopathology. At least one study has demonstrated a role for HSI in the in vitro distinction between normal and lesional skin samples, including melanocytic neoplasms.30 Additionally, this technology has identified differences in spectral signatures of in vivo lesions of cutaneous melanoma along the tumors’ treatment course, potentially supporting a role for HSI in assessing tumor response to chemotherapy.31 The objective of our study was to investigate the potential use of HSI in properly classifying the controversial category of atypical Spitzoid neoplasms. With the identification of differing spectral signatures, HSI could potentially offer much-needed support to this diagnostic dilemma in dermatopathology.

MATERIALS AND METHODS Imaging Spectrometer Hardware An Olympus BX40 upright microscope (Olympus Corporation, Olympus America Inc, Melville, NY) equipped with an H101A ProScan motorized stage (Prior Scientific, Rockland, MD) and ·40 planar objective (numerical aperture = 0.75) was interfaced with the Prism and Reflector Imaging Spectroscopy System (PARISS; LightForm Inc, Asheville, NC). PARISS consists of a curved prism-based imaging spectrometer, a charge-coupled device digital image sensor that measures spectral intensity as a function of wavelength, and an auxiliary digital camera that projects imaged specimens onto a monitor. The curved prism transmits up to 90% of incident light with wavelengths in the visible-to-near infrared light spectrum range of 365–950 nm without second-order pollution and

132

| www.amjdermatopathology.com

results in an increase in overall signal-to-noise ratio when compared with diffraction grating.

Tissue Specimens A total of 102 hematoxylin and eosin (H&E)–stained tissue sections were obtained from the University of Pittsburgh Medical Center Dermatopathology Unit archive upon receiving institutional review board approval (institutional review board no. 0311061; University of Pittsburgh). Each section corresponded to a distinct lesion from various anatomical sites from 102 different patients. Each lesion was classified into 1 of 6 categories (Table 2) according to the final diagnosis rendered in consensus by 3 board-certified dermatopathologists.

Spectral Acquisition Each of the 102 H&E-stained tissue samples was examined with traditional light microscopy for a region of interest (ROI) within the dermis by a board-certified dermatopathologist. An ROI contained a maximal proportion of lesional cells to cellular and noncellular nonlesional tissues. Glass slides were then placed on the microscope stage such that the ROI was centered within the field of view under a magnification of ·400. Forty equal slices across the ROI were imaged by PARISS sequentially and simultaneously for all wavelengths measuring from 385 to 880 nm within a given slice. At a spectral resolution of approximately 1 nm, a spectrum was produced for approximately every 0.6 · 0.6 mm area examined within a sample slice. Such parameters produced 240 spectra per sample, with 496 wavelength data points per spectrum. All measured spectra were calibrated against a National Institute of Standards and Technology traceable lamp.

Spectral Data Analysis The spectrometer generated reflectance data (power [Watts] vs. wavelength [nanometers]) in comma-separated value file format. Each value represented the light intensity recorded by the charge-coupled device digital image sensor for wavelengths spanning 385–880 nm in 1 nm increments. Image cubes of the spectral data, displaying spatial extent with respect to the number of spectral bands in depth, were then created  2013 Lippincott Williams & Wilkins

Am J Dermatopathol  Volume 36, Number 2, February 2014

HSI of Melanocytic Lesions

TABLE 2. Sample Groups

Diagnosis Control: benign melanocytic nevus Control: malignant melanoma Classic Spitz nevus Spitz nevus with unusual features Atypical Spitzoid neoplasm Spitzoid melanoma

Number of Samples (N)

Total Number of Cells

10

5760

576

595

10

4886

488.6

466.5

21

9711

485.55

481.5

20

9177

437

424

20

7795

389.75

384

21

9891

Average Number of Cells

471

Median Number of Cells

431

Cutaneous melanocytic lesions examined in this study distributed among sample groups. The total number of cells within each ROI was manually counted under light microscopy under a ·40 objective.

using ENvironment for Visualizing Images (ENVI; Excelis Inc, Boulder, CO) software. Spectroscopic imaging of each lesion produced 240 spectra for analysis. The clustering technique Iterative Self-Organizing Data Analysis Technique Algorithm (ISODATA)32 was then applied to each of the 6 sample groups, using Euclidean distance as a metric and default values for the number of class inputs. Such an approach reduced the raw data in each sample group into 12 representative spectral classes for more efficient data analysis. The mean spectrum for each of the 12 spectral classes within each of the 6 sample groups was calculated and plotted to better visualize similarities and differences among groups. After initial inspection of the spectral data, a secondorder derivative of each spectrum after continuum removal was performed and function minima that correlated with spectral absorption points were identified. ISODATA was subsequently applied to the entire data set across all sample groups. This analysis was performed in an effort to identify 12 representative spectral classes, whose relative occurrence in any given group may offer points of distinction among melanocytic categories.

FIGURE 1. Spectral plot. Plot of the 240 spectra (wavelength [nanometers] vs. light intensity [Watts]) outputted by PARISS for lesion 1 of 20 from the atypical Spitzoid neoplasm sample group.

that best represented the data set for that particular group. Analysis of the mean spectral vector for each of these 12 representative spectral classes led to the identification of characteristic spectral absorption points for each sample group. Figure 2 displays the mean spectral vector of each of the 12 clustered spectral classes for the atypical Spitzoid neoplasm category. Spectral absorptions for each sample group are listed in Table 3. Examination of these tabulated results lends to certain conclusive remarks. First, all sample groups share common spectral absorptions at 496, 536, and 838 nm. Second, each sample group contains at least one absorption point that is

RESULTS PARISS was used to collect 24,480 spectra from 102 cutaneous melanocytic lesions, which were categorized into 6 distinct groups (Table 2) according to histomorphologic features, in an effort to detect differing spectral characteristics among sample groups. Initial inspection of the data involved plotting the 240 spectra produced per lesion. Figure 1 displays the spectral plot for lesion 1 of 20 from the atypical Spitzoid neoplasm sample group. To efficiently identify any trends, ISODATA was applied to each of the 6 sample groups to yield the 12 spectral classes  2013 Lippincott Williams & Wilkins

FIGURE 2. Plot of mean spectral classes. Plot of the mean spectra of each of the 12 clustered classes that resulted from applying ISODATA to the spectral data for the atypical Spitzoid neoplasm sample group. Absorption points were identified as function minima (3 such points are marked on this plot). www.amjdermatopathology.com |

133

Gaudi et al

Am J Dermatopathol  Volume 36, Number 2, February 2014

TABLE 3. Spectral Absorptions

however, the relative proportion of each class in any given group varies.

DISCUSSION

Spectral absorption points (nanometers of wavelength) for each sample group. Absorption points common to all groups are in bold. Absorption points unique to the respective sample group are encircled.

unique to that group. Third, the Spitzoid melanoma category demonstrates the greatest numbers of total and unique absorption points for a sample group. In subsequent analysis, ISODATA was applied to all sample groups to yield 12 spectral classes representative of the entire data set. As can be seen in Figure 3, each of these 12 spectral classes is present in each of the 6 sample groups;

Our eyes and brain process the wavelength-dependent scatter of visible light photons to characterize and identify objects according to their unique spectral signatures. Modernday imaging spectrometers function with the same underlying principle but are better suited for such a task given their ability to analyze both visible light and near-infrared rays at incremental wavelengths not otherwise perceptible by humans. By acquiring and processing millions of spectra, imaging spectrometers are able to classify a heterogeneous field of view and generate a map showing the precise location of spectra that correlate with target objects. Intrigued by this fundamental concept and encouraged by preliminary data regarding the use of HSI in various areas of medicine, we sought a potential role for this technology in dermatopathology—particularly in the problematic classification of atypical Spitzoid neoplasms. Being the first study of its kind, our initial task was to identify the existence of any spectral characteristics that may be unique to the differing categories of Spitzoid neoplasms. With the discovery of unique spectral signatures, we would substantiate a potential role for HSI in classifying atypical Spitzoid neoplasms and warrant further investigation. Using PARISS HSI instrumentation, we analyzed 102 H&E-stained glass slides corresponding to cutaneous melanocytic lesions categorized into 1 of 6 sample groups (Table 2) on histomorphologic grounds. As listed in Table 1, some of the features lending to a diagnosis of atypical Spitzoid neoplasm were asymmetry, poor circumscription, irregular nesting, deep extension with expansile growth, lack of maturation, and increased cellularity in the dermis with deep mitoses. Thus, these were dermal-predominant lesions—unlike Spitz nevi with unusual features, which were junctional-predominant lesions that contained somewhat greater cytologic atypia and/or pagetoid spread than typically allowed for a diagnosis of classic Spitz nevus. Spitzoid melanomas were those lesions that appeared as classic melanoma at low power

FIGURE 3. Distribution of clustered classes. Bar graph displaying the distribution of the 12 clustered classes that resulted from application of ISODATA to the entire data set spanning all sample groups. Each sample group contains all 12 spectral classes; however, the relative proportions of each spectral class vary in each group.

134

| www.amjdermatopathology.com

 2013 Lippincott Williams & Wilkins

Am J Dermatopathol  Volume 36, Number 2, February 2014

HSI of Melanocytic Lesions

but demonstrated Spitzoid cytology—that is, spindle and epithelioid melanocytes—at high power. Benign melanocytic nevi and malignant melanoma control groups were included in this study to document differences in spectral characteristics at the extremes of lesional tissue. Certain conclusions can be made after reviewing the results of this study. First, similarities in spectral features exist among all sample groups studied—that is, all sample groups share common absorptions at 496, 536, and 838 nm. Second, and more importantly, at least one unique absorption point is noted in each sample group after data clustering within individual sample groups. Similarly, data clustering across all sample groups reveals unique spectral signatures—that is, the 12 spectral classes that result after clustering across the entire data set (24,480 spectra from 102 specimens) reveal the presence of each spectral class in each sample group in varying proportions. Third, Spitzoid melanomas contain the greatest number of total and unique spectral absorptions. This last finding was unforeseen, as the malignant melanoma control group was expected to contain at least as many unique spectra given the tumor’s genetic heterogeneity. A simple explanation for this inconsistency is the small number of specimens comprising the control group. At least one previous study reports significant differences in spectral signatures related to sample preparation.30 We obviated such foreseeable confounders by imaging glass slides that were prepared at a single institution (University of Pittsburgh Medical Center) using an automated, standard H&E staining protocol and identical Leica (Leica Microsystems Inc, Buffalo Grove, IL) glass slides, mounting media, and cover slips. Such an approach presumably limited any variation in sample preparation to day-to-day staining aberrations. Although these preliminary results are encouraging in terms of the potential to further differentiate Spitzoid neoplasms, many questions remain. For example, can we identify spectral classes better suited for distinguishing the melanocytic lesions examined in this study if we alter the manner by which spectral data are clustered? Similarly, will increasing the optical magnification at which slides are spectroscopically imaged lend to superior spectral classes? Should subsequent studies examine multiple ROIs from a given specimen and compare absorption patterns within the same sample? Can we correlate spectral data with molecular data—that is, are some of the observed spectral features attributable to DNA copy number changes or genetic mutations? BRAF mutations are seen in approximately 70% of cases of cutaneous melanoma,33–36 and significant numbers of both benign and dysplastic nevi.33,35–37 Will optical similarities between benign melanocytic nevi and malignant melanoma be associated with known BRAF mutations (or rather the product of a mutated BRAF gene)? Similarly, differences in spectral features between Spitz nevi and malignant melanoma may partially be due to gains of chromosome 11p4–7 or mutations in HRAS4,38 in the former. The results of a recent study lend support to this theory. Through imaging mass spectroscopy, unequivocal cases of Spitz nevi and Spitzoid melanoma were distinguished from one another according to proteonomic differences.39

HSI offers much promise in tissue diagnostics and warrants further investigation. Spectroscopic imaging of H&E-stained glass slides can be simply performed in a matter of minutes at the pathologist’s light microscope, without the need for further tissue processing. In comparison with other ancillary studies, such as immunohistochemistry and molecular techniques, this technology has the potential to offer dramatically superior turnaround times while using significantly less laboratory resources. Although promising in theory and technologically feasible, HSI has a long way to go before it may be implemented in routine laboratory use. Tested sample sizes need to be increased, results validated, and a robust spectral library constructed before this technology can be implemented in routine laboratory use.

 2013 Lippincott Williams & Wilkins

REFERENCES 1. Lodha S, Saggar S, Celebi JT, et al. Discordance in the histopathologic diagnosis of difficult melanocytic neoplasms in the clinical setting. J Cutan Pathol. 2008;35:349–352. 2. Shoo BA, Sagebiel RW, Kashani-Sabet M. Discordance in the histopathologic diagnosis of melanoma at a melanoma referral center. J Am Acad Dermatol. 2010;62:751–756. 3. Barnhill RL, Argenyi ZB, From L, et al. Atypical Spitz nevi/tumors: lack of consensus for diagnosis, discrimination from melanoma, and prediction of outcome. Hum Pathol. 1999;30:513–520. 4. Bastian BC, LeBoit PE, Pinkel D. Mutations and copy number increase of HRAS in Spitz nevi with distinctive histopathological features. Am J Pathol. 2000;157:967–972. 5. Bastian BC, Olshen AB, LeBoit PE, et al. Classifying melanocytic tumors based on DNA copy number changes. Am J Pathol. 2003;163: 1765–1770. 6. Bastian BC, Wesselmann U, Pinkel D, et al. Molecular cytogenetic analysis of Spitz nevi shows clear differences to melanoma. J Invest Dermatol. 1999;113:1065–1069. 7. Harvell JD, Kohler S, Zhu S, et al. High-resolution array-based comparative genomic hybridization for distinguishing paraffin-embedded Spitz nevi and melanomas. Diagn Mol Pathol. 2004;13:22–25. 8. Lillesand TM, Kiefer RW, Chipman JW. Remote Sensing and Image Interpretation. New York, NY: John Wiley & Sons Inc; 2003. 9. McGoverin CM, Engelbrecht P, Geladi P, et al. Characterisation of nonviable whole barley, wheat and sorghum grains using near-infrared hyperspectral data and chemometrics. Anal Bioanal Chem. 2011;401:2283–2289. 10. Sivertsen AH, Heia K, Stormo SK, et al. Automatic nematode detection in cod fillets (Gadus morhua) by transillumination hyperspectral imaging. J Food Sci. 2011;76:S77–S83. 11. Franch-Lage F, Amigo JM, Skibsted E, et al. Fast assessment of the surface distribution of API and excipients in tablets using NIR-hyperspectral imaging. Int J Pharm. 2011;411:27–35. 12. Yudovsky D, Nouvong A, Pilon L. Hyperspectral imaging in diabetic foot wound care. J Diabetes Sci Technol. 2010;4:1099–1113. 13. Yudovsky D, Nouvong A, Schomaker K, et al. Monitoring temporal development and healing of diabetic foot ulceration using hyperspectral imaging. J Biophotonics. 2011;4:565–576. 14. Yudovsky D, Nouvong A, Schomaker K, et al. Assessing diabetic foot ulcer development risk with hyperspectral tissue oximetry. J Biomed Opt. 2011;16:026009. 15. Chin JA, Wang EC, Kibbe MR. Evaluation of hyperspectral technology for assessing the presence and severity of peripheral artery disease. J Vasc Surg. 2011;54:1679–1688. 16. Zakian C, Pretty I, Ellwood R. Near-infrared hyperspectral imaging of teeth for dental caries detection. J Biomed Opt. 2009;14:064047. 17. Johnson WR, Wilson DW, Fink W, et al. Snapshot hyperspectral imaging in ophthalmology. J Biomed Opt. 2007;12:014036. 18. Lee N, Wielaard J, Fawzi AA, et al. In vivo snapshot hyperspectral image analysis of age-related macular degeneration. Conf Proc IEEE Eng Med Biol Soc. 2010;2010:5363–5366.

www.amjdermatopathology.com |

135

Gaudi et al

Am J Dermatopathol  Volume 36, Number 2, February 2014

19. Medina JM, Pereira LM, Correia H, et al. Hyperspectral optical imaging of human iris in vivo: characteristics of reflectance spectra. J Biomed Opt. 2011;16:076001. 20. Nourrit V, Denniss J, Muqit MM, et al. High-resolution hyperspectral imaging of the retina with a modified fundus camera. J Fr Ophtalmol. 2010;33:686–692. 21. Sohrab MA, Smith RT, Fawzi AA. Imaging characteristics of dry agerelated macular degeneration. Semin Ophthalmol. 2011;26:156–166. 22. Best SL, Thapa A, Holzer MJ, et al. Minimal arterial in-flow protects renal oxygenation and function during porcine partial nephrectomy: confirmation by hyperspectral imaging. Urology. 2011;78:961–966. 23. Holzer MS, Best SL, Jackson N, et al. Assessment of renal oxygenation during partial nephrectomy using hyperspectral imaging. J Urol. 2011; 186:400–404. 24. Panasyuk SV, Yang S, Faller DV, et al. Medical hyperspectral imaging to facilitate residual tumor identification during surgery. Cancer Biol Ther. 2007;6:439–446. 25. Zuzak KJ, Naik SC, Alexandrakis G, et al. Intraoperative bile duct visualization using near-infrared hyperspectral video imaging. Am J Surg. 2008;195:491–497. 26. Ferris DG, Lawhead RA, Dickman ED, et al. Multimodal hyperspectral imaging for the noninvasive diagnosis of cervical neoplasia. J Low Genit Tract Dis. 2001;5:65–72. 27. Martin ME, Wabuyele MB, Chen K, et al. Development of an advanced hyperspectral imaging (HSI) system with applications for cancer detection. Ann Biomed Eng. 2006;34:1061–1068. 28. Akbari H, Uto K, Kosugi Y, et al. Cancer detection using infrared hyperspectral imaging. Cancer Sci. 2011;102:852–857.

29. Siddiqi AM, Li H, Faruque F, et al. Use of hyperspectral imaging to distinguish normal, precancerous, and cancerous cells. Cancer. 2008 25;114:13–21. 30. Dicker DT, Lerner J, Van Belle P, et al. Differentiation of normal skin and melanoma using high resolution hyperspectral imaging. Cancer Biol Ther. 2006;5:1033–1038. 31. Dicker DT, Kahn N, Flaherty KT, et al. Hyperspectral imaging: a noninvasive method of imaging melanoma lesions in a patient with stage IV melanoma, being treated with a RAF inhibitor. Cancer Biol Ther. 2011 15;12:326–334. 32. Tou JT, Gonzalez RC. Pattern Recognition Principles. Boston, MA. Addison-Wesley Educational Publishers Inc; 1977. 33. Ibrahim N, Haluska FG. Molecular pathogenesis of cutaneous melanocytic neoplasms. Annu Rev Pathol. 2009;4:551–579. 34. Meyle KD, Guldberg P. Genetic risk factors for melanoma. Hum Genet. 2009;126:499–510. 35. Palmieri G, Capone M, Ascierto ML, et al. Main roads to melanoma. J Transl Med. 2009;7:86. 36. Sekulic A, Haluska P Jr, Miller AJ, et al. Malignant melanoma in the 21st century: the emerging molecular landscape. Mayo Clin Proc. 2008;83: 825–846. 37. Pollock PM, Harper UL, Hansen KS, et al. High frequency of BRAF mutations in nevi. Nat Genet. 2003;33:19–20. 38. van Engen-van Grunsven AC, van Dijk MC, Ruiter DJ, et al. HRASmutated Spitz tumors: a subtype of Spitz tumors with distinct features. Am J Surg Pathol. 2010;34:1436–1441. 39. Lazova R, Seeley EH, Keenan M, et al. Imaging mass spectrometry— a new and promising method to differentiate Spitz nevi from Spitzoid malignant melanomas. Am J Dermatopathol. 2012;34:82–90.

136

| www.amjdermatopathology.com

 2013 Lippincott Williams & Wilkins

Hyperspectral imaging of melanocytic lesions.

Hyperspectral imaging (HSI) allows the identification of objects through the analysis of their unique spectral signatures. Although first developed ma...
446KB Sizes 0 Downloads 0 Views