In Vivo Measurement and Modeling of Multispectral Reflectance Images for Melanoma Diagnosis

Abstract

Listen

Translate article

Noninvasive detection of malignant melanoma in early stages is critical to improve patients’ prognosis. We acquired in vivo reflectance images of dysplastic lesions from 12 patients at 31 wavelengths from 500 to 950 nm. Based on these image data, we developed a parallel Monte Carlo code to simulate reflectance images from a heterogeneous skin tissue model. With this tool, we have investigated the dependence of the lesion contrast in the reflectance image on the heterogeneous distribution of tissue optical parameters. The Monte Carlo model is currently used to generate multispectral reflectance imaging data for multivariate analysis of the in vivo imaging data. DOI: 10.2529/PIERS061020122702 It has been estimated that Caucasians may develop up to 50 clinically benign nevi by age 40. Patients with more than 100 nevi were estimated to have a 3-fold to 10-fold increased risk of developing malignant melanomas and pigmented basal cell carcinomas in comparison to the general population [1]. Diagnosis of MM is currently established by histopathology of biopsied tissues from the suspicious-appearing nevi or pigmented lesions. These patients often present a difficult dilemma to primary-care physicians and dermatologists. A physician has to either prescribe painful and costly excision biopsy with likely cosmetic disfigurement with limited information on the lesion or leave untouched with the risk of MM developing in the patients. Therefore, cost-effective pre-biopsy methods of examination could greatly improve patient care and reduce medical cost with better specificity and sensitivity than what are available now. In this report, we present multispectral reflectance image data acquired from 12 patients with dysplastic lesions at 31 wavelengths from 500 to 950 nm and results of numerical studies of reflectance imaging method by a Monte Carlo (MC) code. A multispectral imaging system employing a thermoelectrically cooled CCD camera has been constructed to acquire polarimetric images [2]. Fig. 1 presents a schematic of the imaging system. We used a xenon fiber optic light source and a collimating lens to produce a parallel light beam of