Sections from human MM biopsies were acquired from the University of Szeged, Hungary with full ethical approval from the host institute and patient consent. Tissue devitalization protocols were standardized for all patients and no samples showed any signs of necrosis prior to excision. Histological preparation of the tissue was identical for all samples. Five samples were diagnosed as superficial spreading MM, one nodular MM and two lentigo mailgnia MM. Three samples were from female patients and five from males with an age range of 58–75 years. Breslow thickness assessments ranged from 0.304 to 1.672 mm and post-MM excision survival was 30–84 months.
Multiple adjacent fields of unstained human melanoma skin sections (4µm) (n
= 8 patients) were imaged using SHG imaging to reveal fibrillar collagen distribution (principally Type I and III), followed by single-photon transmission (bright field) imaging. A detailed description of the experimental setup is described in [19
]. Image acquisition was performed on a Leica SP2 AOBS confocal multi-photon laser scanning microscope coupled to a Leica DMRE upright microscope (Leica, Milton Keynes, UK), equipped with a pulsed, mode locked femtosecond (fs) Ti:sapphire Tsunami laser synchronously pumped by a Millenia VII (Spectra-Physics, Mountain View CA), diode pumped solid state frequency doubled laser capable of delivering up to 8.5W pumping power at 532nm.
All SHG image acquisition was performed at 880nm, which has previously been found to be the optimum wavelength for collagen types I and III [19
]. The pulse width of the Tsunami was 80 fs with a pulse repetition rate of 80 Mhz. Laser power output at the microscope objective was recorded with a coherent power meter and calibrated for all samples to deliver peak power of 0.4 × 108
and was consistently maintained below the damage threshold of the samples, which for collagen in sections was found to be 1.5 × 108
]. An IST laser spectrum analyzer coupled to a Tektronix TDS 210 oscilloscope was used to tune the laser to the desired wavelength. An electro-optical modulator (EOM) (Linos LIV20) received the laser output before delivery to the confocal microscope through a series of optical mirrors. The EOM allowed the laser intensity at the objective to be controlled and optimized. EOM was set at 90% to circularly polarize the pump laser beam, so as to ensure the incident beam laser remained consistent across all specimens. Thus the potential effects of sample polarization versus fundamental pump beam polarization were minimized.
In order to image the complete section we used a motorized stage (Mietzhouser, GmbH), controlled by the Leica software to perform automated montages from multiple fields consisting of up to 42 x 19 fields with up to 10 z-sections of 1µm thickness. Samples were imaged with a 20x, 0.7NA dry objective and 2x zoom producing an image measuring 375 by 375 µm at 512 by 512 pixel resolution. The SHG signal in the backscattered geometry was captured in the de-scanned detector with the pinhole set to the maximum of 600 µm. The Leica microscope incorporates a programmable, prismatic beam splitter, which is capable of single and multi-chromatic beam splitting. The backscattered photons are subsequently focused through the objective lens with an estimated beam diameter of 0.24 µm (λp
= 880 nm) and delivered through the prismatic beam splitter, programmed to collect light at 435-445nm, before passing it on to the photomultiplier tube (PMT). This calculation is based on the Guoy phase shift and its relationship to the NA of the objective lens, the fundamental wavelength and refractive index of the sample [26
]. Thus the beam diameter = 0.32λ / √2NA. The coherent SHG signal formed in the transmission geometry was detected by the transmission detector via the microscope condenser with 445 nm df 30nm band pass filter inserted in the light path (Chroma Inc. UK). In addition, a non-confocal, bright-field image was acquired (sequentially to the SHG image) using single photon excitation at 488nm, which was collected in the transmission PMT. This image served as a reference image to overlay the H&E or Melan-A with the SHG images during the assessment stage. At the end of image acquisition the automatic montage software produced a final compilation image of the all z
Sister-sections were then imunohistochemically stained for the melanoma marker, Melan-A, and imaged using conventional light microscopy, then overlaid on the SHG images for comparison. The extent of invasion and limits of the borders between the two imaging techniques (H&E and or Melan-A) was assessed and marked by two independent histopathologists. The lateral boarders of cancer invasion were marked on the SHG images by an independent researcher, blinded to the histopathologists assessments, then checked for the degree of correlation between the SHG and H&E/Melan-A findings, using Mann-Whitney U-Test and Kendall-Rank Correlation Analysis. The criteria used for the independent evaluation of the MM borders of the SHG images were that one could clearly discriminate undisrupted collagen fiber bundles at the dermal-epidermal junction, adjacent to the MM lesion.
Collagen fiber density was quantified from seven regions using ImageJ image analysis software (Freeware, rsbweb.nih.gov
) to calculate the percentage area of collagen fibers for that region. The images were segmented using an adaptive threshold algorithm with a fixed kernel size and offset for all samples. Data was expressed as Area %, calculated using area of collagen fibers/ area of measuring frame x 100. Generally SHG produces very high contrast images with very low or no background and since pump laser beam polarization was circular, segmentation based on threshold values proved to be a robust method. The regions quantified are highlighted in
and were an area of non-lesion skin under the dermal-epidermal junction (labeled A
), the lateral melanoma border under dermal-epidermal junction (labeled B
), an area under the middle part of the lesion and under the dermal-epidermal junction (labeled C
), the areas labeled D, E, F,
are all below C
but one field vertically deeper to each other, and the area of dermis at the same depth as F
but away from the lesion (labeled G
Fig. 2 (A) Enlarged view of showing regions from which collagen fiber density was quantified. A = Area of normal skin under dermal-epidermal junction, B = Lateral melanoma border under dermal-epidermal junction, C = Area under the middle part of the (more ...)
Fresh pieces of whole rat and human skin were cleared of subcutaneous fat and connective tissue, and hair in the case of the rat samples, and mounted onto glass slides in PBS buffer. Human skin samples were collected from The Royal London Hospital Trust, department of surgery with full patient consent and ethical approval, and imaged within four hours. One sample was from 22 year old, breast reduction patient, and the other three samples were breast reconstructions from age rages of 30–45 years. Two samples appeared heavily pigmented whilst the other two appeared very lightly pigmented. SHG image acquisition was performed in both backscattered and forward propagated (transmission) geometries from the epithelial surface and imaged at full depth (deep into the dermis) in order to demonstrate that the technique is transferable to intact live skin. A Leica 20x 0.7 NA, long working distance objective, was used for image acquisition. A single field of 512 x 512 pixels (750 x 750 µm or 430 x 430 µm) at zoom 1.0 or 1.7 at 800 Hz with 2.5µm z-step, an imaging average of 2 frames was acquired for both rat and human ex vivo samples. Typical imaging times for ex vivo skin samples were 0.7 s per frame and up to 225 z-slices per sample; therefore total-imaging times did not exceed 3 min. 3D-images were generated with Imaris visualization software (Bitplane Scientific Software. Zurich, Switzerland).