Distinguishing between benign and malignant melanocytic nevi by in vivo multiphoton microscopy.

MPTflex clinical tomograph The laser-scanning based clinical multiphoton tomograph, MPTflex (JenLab GmbH, Germany) consists of a compact, turn-key femtosecond laser (MaiTai Ti:Sappire oscillator, sub-100 fs, 80 MHz, tunable 690–1020 nm; Spectra Physics, Mountain View, CA), an articulated arm with near-infrared optics, and beam scanning module. The system has two photomultiplier tube (PMT) detectors employed for parallel acquisition of TPEF and SHG signals. A customized metallic ring taped on the subject’s skin attaches magnetically to the objective holder in the articulated arm, minimizing motion artifacts. The excitation wavelength used for this study was 790 nm. The TPEF signal was detected over the spectral range of 410nm–650nm while the SHG signal was detected over a narrow spectral bandwidth 385nm–405nm through emission filters placed in the TPEF and SHG detection channels, respectively. We used a Zeiss objective (40X, 1.3NA, oil immersion) for focusing into the tissue.

Study design

We imaged 15 melanocytic nevi (5 common nevi, 5 dysplastic nevi and 5 melanoma) in 14 patients. All in vivo measurements were conducted according to an approved institutional protocol with written informed consent obtained from all patients. The 15 lesions were distributed in 6 primary locations for all patients, including back ( 4 ), arms ( 4 ), legs ( 1 ), chest ( 1 ), abdomen ( 3 ) and face ( 2 ). MPM measurements were performed on lesion sites as well as on adjacent normal skin. Optical sections of about 200×200 μm 2 at different depths ranging from 0 to about 200 μm (5 μm steps) were obtained. The time required for each optical section was 6 s. As the optical section is limited to a small scan field, the overall investigation of the lesion required the acquisition of several image stacks of different skin sites. We acquired about three image stacks for each lesion. All lesions clinically diagnosed as dysplastic nevi and melanoma by board certified dermatologists (KMK and CBZ) were biopsied and diagnosed by a dermatopathologist (RMH), using standard H&E histopathology. For the qualitative analysis, we compared MPM and histologic images in order to determine whether H&E histopathology hallmarks could be correlated with structures in in vivo MPM images. Quantitative methods are described below.

Image analysis

All images were processed using ImageJ ( 15 ). For the quantitative analysis we wrote macros for automatic measurement of key parameters characteristic of TPEF and SHG images. TPEF images were also used to identify and calculate the number of melanocytic dendrites. A composite “multi-photon melanoma index (MMI)” was linearized on a 0–9 scale. For the TPEF contrast we used the raw TPEF images to calculate the ratio (F) between the spatial standard deviation and the mean pixel intensity for each TPEF image in a z-stack corresponding to the epidermal-dermal junction (EDJ). F = σ F l < I > where σ Fl is the spatial standard deviation of fluorescence signal intensity and the mean pixel intensity in the TPEF image. For each lesion, F ratios were calculated over 5 consecutive image planes (spanning a total of 20 μm in depth), starting with the image of the basal layer in each acquired stack and going down into the EDJ. We considered the EDJ starting at the location where the collagen structure in the top of the papilla was visualized. The basal layer was defined as the first cell layer above the EDJ. A mean F value was calculated for each lesion. The mean represents the average F value over the TPEF images analyzed in all stacks of the lesion. Parameter F is related to epidermal features assessed qualitatively by histopathology, such as lentiginous hyperplasia and Pagetoid spread. By its definition, it measures the degree of pixel intensity homogeneity in the MPM image. For the SHG contrast , the SHG images were converted to 8-bit images and subsequently to binary images by using the automatic thresholding function in ImageJ ( 16 ). The automatic thresholding procedure was suitable for our analysis because the signal-to-background ratio (SBR) of the SHG images was high. The average SBR for each of the SHG images included in the analysis was at least 10:1 and typically averaged greater than 20:1. The “bright pixels” were defined as the pixels of value 1 in the binary images. We defined the density of the bright pixels by the ratio between the number of bright pixels and total number of pixels in one image. The density of bright pixels was calculated for each image in a z-stack of 8 consecutive image planes (spanning a total of 35 μm in depth), starting with the first SHG image of the EDJ. For each stack, a parameter S was defined as: S = σ SHG < ρ > where σ SHG is the standard deviation and the mean density of bright pixels in the binary SHG images of the stack. A mean S value was calculated for each lesion. The mean represents the average S value over all the stacks of SHG images of the lesion. Parameter S , by its definition, is a measurement of the change in collagen across the EDJ and therefore, a measurement of histopathology features such as irregular nests of nevus cells along the basal layer, erosion of the junction, and invasion of melanocytes into the dermis. A large S reflects a rapid increase in the collagen amount from the top of the dermal papillae to deeper layers in the papillary dermis. A small S value reflects a slower increase in collagen content across the junction, which is due to the presence of cells in the papillary dermis; typically nevus cells from the sides of the rete ridges in dysplastic nevi and melanoma cells in melanoma lesions. A larger volume at the EDJ would comprise more information for image analysis, but for dark, highly pigmented nevi, TPEF and SHG signals diminish with depth due to high absorption and scattering. Stacks of 20 μm and 35 μm total thickness for TPEF and SHG images respectively, proved to contain relevant information in all analyzed lesions. To determine melanocytic dendrite density , we used the NeuronJ plug-in ( 17 ) in ImageJ for tracing and counting of melanocytic dendrites in the TPEF images corresponding to spinosum and granulosum epidermal layers. We calculated a density parameter D, which was defined as the number of melanocytes in a stack volume. D = N V where N is the number of melanocytic dendrites in the stratum spinosum and granulosum of the epidermis and V is the volume (the image area x the thickness of the epidermis from the stratum corneum to the basal layer). A mean D value was calculated for each lesion. The mean represents the average D value obtained from melanocytic dendrites counted by two independent observers. Parameter D represents a measurement of the density of melanocytic dendrites in upper epidermal layers. A high density number is a hallmark of melanoma, but a limited number of melanocytic dendrites is allowed in the stratum spinosum of the epidermis in dysplastic nevi ( 18 ). Parameter D was introduced in order to address this ambiguity.

Supplementary Material 1 2 3