Photothermal Microscopy and
Spectroscopy Webinar

Malignant melanoma (MM) is one of the most common cancers worldwide. It has a favorable prognosis only if the affected area is removed at an early stage. MM reportedly causes the large majority of skin cancer deaths despite the fact that it accounts for < 2% of skin cancer cases [1]. The incidence of MM has been increasing for > 30 years and one of its most ominous characteristics is its high propensity to produce distant metastases, because it can get disseminated throughout the body through lymphatic and hematogenous spread. For this reason, early detection and treatment of MM are crucial life-saving measures [2]. Although dermoscopy is a powerful diagnostic technique [3] and the ABCDE (abbreviation for asymmetrical shape, border, color, diameter, and evolution) rule provides a guide to the identification of involved areas, pathological examination is the gold standard for MM diagnosis. But diagnosis remains highly reliant on the skill level of the pathologist.

We developed a photothermal microscope with resolution beyond diffraction limit [4]. The textural structure of the images of the mouse skin samples containing benign epithermal tumor (Pa), carcinoma (Ca), and metastatic carcinoma (Mc) cells taken with the PT imaging method were analyzed by GLCM. The nine parameters were calculated as shown below. The areas of imaging data were analyzed by the GLCM method, and the 8-bit level gray level intensity distribution of the PT signal was obtained. Twelve images of 72 × 72 μm ² to multiple samples of excitation at 488 nm for both nevus and MM samples were also obtained. The areas of imaging data were analyzed by this method, and the 8-bit level gray level intensity distribution of the PT signal was analyzed using the GLCM analysis. There were total of 48 images with an area of 18 × 18 μm ² from both nevus and MM providing sufficient data of various parts of skin. We found that a few of the nine GLCM parameters clearly showed the ability to discriminate between nevus and MM, and can hopefully be used as criteria for pathological diagnosis [4,5].

We then performed receiver operating characteristic (ROC) curve analysis based on fitted Gaussian curves to the observed distribution. An ROC curve is commonly used to evaluate the diagnostic ability of a test. When a threshold parameter used in the system classifying examinees into two groups, positive and negative for some features, this curve is plotted as the sensitivity against the false positive ratio. The d parameter d = 10 provides the best performance for all nine GLCM parameters in both cases pumped at 405 nm and pumped at 488 nm. We plotted ROC curves for nine parameters based. The area under the curve (AUC) is an indicator of the diagnostic ability; > 0.9, 0.7 ~ 0.9, and < 0.7 correspond to high accuracy, moderate accuracy and poor accuracy, respectively. Entropy, Contrast and Variance show high AUCs, namely 0.909, 0.905 and 0.897, respectively. These values indicate that those parameters provide highly accurate methods to distinguish nevus and MM cells. In the case of 488 nm excitation, the AUCs of Prominence, Variance and Shade are 0.812, 0.808 and 0.768, respectively, indicating not as good performance as 405 nm excitation [4,5]. We also discussed the image features of the three types (Pa, Ca, and Mc) of cells and also in terms of fractal dimensions.

1. American Cancer Society, Melanoma Skin Cancer Overview (2015).
2. D. S. Rigel, J. A. Carucci, CA Cancer J. Clin. 50, 215–236 (2000).
3. H. P. Soyer, J. Smolle, H. Kerl, H. Stetner, Lancet 2, 803 (1987).
4. T. Kobayashi et al., Photonics (MDPI) 25, 271 (2018).
5. T. Kobayashi, Bioengineering. (MDPI) 5, 67 (2018).