Abstract
Pump-probe microscopy is an emerging molecular imaging technique that probes the excited state dynamics properties of pigmented samples. This method has been particularly intriguing for melanoma because, unlike other methods available, it can provide nondestructive, quantitative chemical information regarding different types of melanins, with high spatial resolution. In this Letter, we present a method based on mathematical morphology to quantify melanin structure (eumelanin, pheomelanin, and total melanin content, uniquely available with pump-probe microscopy) to aid in melanoma diagnosis. The approach applies a two-dimensional autocorrelation function and utilizes statistical parameters of the corresponding autocorrelation images, specifically, the second moments and entropy, to parameterize image structure. Along with bulk melanin chemical information, we show that this method can differentiate invasive melanomas from noninvasive and benign lesions with high sensitivity and specificity (92.3% and 97.5%, respectively, with N=53). The mathematical method and the statistical analysis are described in detail and results from cutaneous and ocular conjunctival melanocytic lesions are presented.