Open AccessIncorporating patient demographics into Raman spectroscopy algorithm improves in vivo skin cancer diagnostic specificity
Author(s) -
Zhao Jianhua,
Zeng Haishan,
Kalia Sunil,
Lui Harvey
Publication year - 2019
Publication title -
translational biophotonics
Language(s) - English
Resource type - Journals
ISSN - 2627-1850
DOI - 10.1002/tbio.201900016
Subject(s) - basal cell carcinoma , medicine , demographics , actinic keratosis , skin cancer , seborrheic keratosis , dermatology , lesion , receiver operating characteristic , basal cell , in vivo , cancer , oncology , pathology , biology , demography , sociology , microbiology and biotechnology
The study objective is to evaluate whether incorporating patient demographics into Raman spectral analysis can improve diagnostic performance. In vivo Raman spectra of 731 cases were analyzed by dividing the data into two groups: skin cancers/precancers (malignant melanoma, basal cell carcinoma, squamous cell carcinoma, and actinic keratosis, n = 340) and benign lesions (pigmented nevi and seborrheic keratosis, n = 391). Patient age, gender, skin type and location of the lesion were taken into account in the analysis. Multivariate statistical analysis including principal component and general discriminant analysis and partial least squares (PLS) were utilized for lesion discrimination. Based on PLS analysis, the area under receiver operating characteristic curve was improved from 0.913 to 0.934 ( P < .05) after incorporating patient demographics into the algorithm; the specificity was increased from 33.5% to 44.5%, 56.0% to 68.5% and 76.0% to 82.1% for sensitivity of 99%, 95% and 90%, respectively ( P < .05 for all sensitivity levels).