Dynamic infrared imaging for skin cancer screening

• Published in: Infrared physics & technology 2015-05, Vol.70, p.147-152
• Main Authors: Godoy, Sebastián E., Ramirez, David A., Myers, Stephen A., von Winckel, Greg, Krishna, Sanchita, Berwick, Marianne, Padilla, R. Steven, Sen, Pradeep, Krishna, Sanjay
• Format: Article
• Language: English
• Subjects: Dynamic thermal imaging; Infrared imaging; Non-invasive screening; Skin cancer; Skin cancer screening
• ISSN: 1350-4495; 1879-0275
• DOI: 10.1016/j.infrared.2014.09.017

•We discussed and proposed a standardized analysis method for dynamic thermal imaging of actual patient data.•We observed that selecting pixels with the same initial temperature is the key enabling tool in the analysis of the data.•We extensively tested the methodology on more than 100 human subjects (ClinicalTrials ID number NCT02154451).•We achieved a sensitivity of 95%, (95% CI: [87.8% 100.0%]), and a specificity of 83%, (95% CI: [73.4% 92.5%]). Dynamic thermal imaging (DTI) with infrared cameras is a non-invasive technique with the ability to detect the most common types of skin cancer. We discuss and propose a standardized analysis method for DTI of actual patient data, which achieves high levels of sensitivity and specificity by judiciously selecting pixels with the same initial temperature. This process compensates the intrinsic limitations of the cooling unit and is the key enabling tool in the DTI data analysis. We have extensively tested the methodology on human subjects using thermal infrared image sequences from a pilot study conducted jointly with the University of New Mexico Dermatology Clinic in Albuquerque, New Mexico (ClinicalTrials ID number NCT02154451). All individuals were adult subjects who were scheduled for biopsy or adult volunteers with clinically diagnosed benign condition. The sample size was 102 subjects for the present study. Statistically significant results were obtained that allowed us to distinguish between benign and malignant skin conditions. The sensitivity and specificity was 95% (with a 95% confidence interval of [87.8% 100.0%]) and 83% (with a 95% confidence interval of [73.4% 92.5%]), respectively, and with an area under the curve of 95%. Our results lead us to conclude that the DTI approach in conjunction with the judicious selection of pixels has the potential to provide a fast, accurate, non-contact, and non-invasive way to screen for common types of skin cancer. As such, it has the potential to significantly reduce the number of biopsies performed on suspicious lesions.