Optical delineation of human malignant melanoma using second harmonic imaging of collagen

Abstract

Skin cancer incidence has increased exponentially over the last three decades. In 2008 skin cancer caused 2280 deaths in the UK, with 2067 due to malignant melanoma. Early diagnosis can prevent mortality, however, conventional treatment requires multiple procedures and increasing treatment times. Second harmonic generation (SHG) imaging could offer diagnosis and demarcation of melanoma borders non-invasively at presentation thereby short-cutting the excision biopsy stage. To test the efficacy and accuracy of SHG imaging of collagen in skin and to delineate the borders of skin cancers, unstained human melanoma biopsy sections were imaged using SHG microscopy. Comparisons with sister sections, stained with H&E or Melan-A were made for correlation of invasion borders. Fresh ex vivo normal human and rat skin was imaged through its whole thickness using SHG to demonstrate this technique is transferable to in vivo tissues. SHG imaging demonstrated detailed collagen distribution in normal skin, with total absence of SHG signal (fibrillar collagen) within the melanoma-invaded tissue. The presence or absence of signal changes dramatically at the borders of the melanoma, accurately demarcating the edges that strongly correlated with H&E and Melan-A defined borders (p<0.002). SHG imaging of ex vivo human and rat skin demonstrated collagen architecture could be imaged through the full thickness of the skin. We propose that SHG imaging could be used for diagnosis and accurate demarcation of melanoma borders on presentation and therefore potentially reduce mortality rates.

Keywords: (170.3880) Medical and biological imaging; (180.1790) Confocal microscopy; (180.4315) Nonlinear microscopy; (190.1900) Diagnostic applications of nonlinear optics.

Fig. 2

(A) Enlarged view of Fig. 1D 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 lesion and under dermal-epidermal junction, areas D, E, F, are all below C but one field vertically deeper to each other, area G is at the same depth as F but away from the lesion (A). (B) Graph of collagen fiber density (Area %, ± SE) of the individual regions of A (n = 5-8). On some samples not all regions could be quantified because the biopsies were cut to shallow, i.e., not enough dermis depth present, therefore the deeper regions were absent and hence there is no data and the n-number vary. Note regression line illustrating a linear correlation of collagen fiber density (R² = 0.9849) from region C (mid-point of MM lesion) to region G (deep area of non-lesion skin). Scale bar = 1mm

Fig. 1

Montage of SHG images in the forward propagated (A), backscattered geometry (B), and bright field image (C). Superimposed images of bright-field and SHG images indicate collagen distribution within each section (D). Scale bar = 1mm

Fig. 3

Cross-reference control comparison of melanoma borders obtained from H&E (A) and Melan-A stained sister sections (B) and comparison with SHG (C) showed a high degree of correlation (p<0.0012). Vertical lines (black lines in A & B; white lines in C) represent lateral borders of MM as identified by two independent histopathologists (A and B) and independent researcher (C). Red box (B-C) represents the areas shown at high power in Figs. 4A and 4B, respectively. Scale bar = 1mm

Fig. 4

High magnification images of Fig. 3. Melan-A staining (A), SHG/brightfield image (B), corresponding to left hand border (red boxes) in Figs. 3B and 3C, respectively. Note that the SHG signal starts to reappear at the dermo-epidermal junction at the point where MM lesion is absent. Scale bar = 1mm

Fig. 5

Measured distance from center of each MM sample to left and right borders of lesion for SHG (dotted infill = left border, horizontal line infill = right border) stained section and H&E (diagonal hatch infill = left border, checker hatch infill = right border) imaged sister section (A). Mean difference of distance for left and right MM borders of H&E versus SHG images (B). Statistical analysis shows no significant difference between the two methods with total p-value for both sides p = 0.9339, whilst Kendall-Rank Correlation Analysis Showed a total p-value for both sides p = 0.0012.

Fig. 6

SHG imaging of normal, ex vivo, live rat skin (A) and human skin sample (B) showing collagen morphology throughout the entire thickness when imaged from the epidermis, in the transmission (red) and back-scattered (green) geometry. Rat skin, SHG transmission images (red) and back-scattered SHG signal (green) were detected through the whole thickness of the skin (A). The blue channel is the two-photon auto fluorescence image of the upper epidermis, whilst the yellow represents the colocalization of the red and green channels. In lightly pigmented human skin samples, we were able to image through the whole thickness of epidermis and dermis to depths of approximately 300µm in the back-scattered geometry (green signal) and over 1000µm in the transmission geometry (red signal) (B). Scale bars = 50 µm for each orthogonal plane.