Open AccessDiffuse reflectance spectroscopy as a potential method for nonmelanoma skin cancer margin assessment
Author(s) -
Zhang Yao,
Moy Austin J.,
Feng Xu,
Nguyen Hieu T. M.,
Sebastian Katherine R.,
Reichenberg Jason S.,
Markey Mia K.,
Tunnell James W.
Publication year - 2020
Publication title -
translational biophotonics
Language(s) - English
Resource type - Journals
ISSN - 2627-1850
DOI - 10.1002/tbio.202000001
Subject(s) - medicine , basal cell carcinoma , skin cancer , receiver operating characteristic , basal cell , margin (machine learning) , logistic regression , dermatology , pathology , radiology , cancer , oncology , computer science , machine learning
The standard‐of‐care for nonmelanoma skin cancer (NMSC), including basal cell carcinoma (BCC) and squamous cell carcinoma (SCC), is surgical excision of the tumor. To ensure removal of the entire tumor and minimize removal of healthy tissue, a tool for quick discrimination between cancerous and normal skin is needed. We propose using diffuse reflectance spectroscopy (DRS) to assess tumor margins. We collected two independent clinical datasets. Using a clinical study of 31 patients with 50 NMSC lesions, we trained models including DRS data normalization, a physiological model, parameter selection and logistic regression classifiers. We tested our models on a second clinical dataset of 56 patients with 68 NMSC lesions. Our test results showed the area under the ROC curve of 0.94 for BCC vs normal and 0.90 for SCC vs normal. When applying a threshold selected using the training dataset to the test dataset, for BCC vs normal, specificity is 89% and sensitivity is 87%; for SCC vs normal, specificity is 65% and sensitivity is 97%. This study indicates that DRS can be potentially used to map the tumor margin prior to surgery and monitor margins during the surgery on the surface of the skin.