BACKGROUND: The accuracy of diagnosis of skin cancer and especially of early malignant melanoma is most important to reduce its morbidity and mortality. Previous pilot studies using electrical impedance measurements indicate statistically significant accuracies for the detection of skin cancer. OBJECTIVES: The aim of this study is to investigate the accuracy of electrical impedance spectra to distinguish between malignant melanoma and benign skin lesions using an automated classification algorithm. PATIENTS/METHODS: Electrical impedance spectra were measured in a multi-centre study at 12 clinics around Europe. Data from 285 histologically analysed lesions were used to train an algorithm to sort out lesions for automatic detection of melanoma. Another data cohort of 210 blinded lesions (148 various benign lesions and 62 malignant melanomas where 38 being from Breslow thickness ≤1 mm) from 183 patients was thereafter used to estimate the accuracy of the technique. RESULTS: Observed sensitivity to malignant melanoma is 95% (59/62) and observed specificity 49% (72/148). CONCLUSIONS: The results suggest that electrical impedance spectra can distinguish between malignant melanoma and benign skin lesions. Although it is indicated that the accuracy of the device is clinically promising, the overall performance, and the sensitivity to thin malignant melanomas, must be improved and thoroughly validated before the instrument can be used as a routine stand-alone diagnostic decision support tool. The technique is under revision to further improve the reproducibility, specificity and sensitivity.

Division of Imaging and Technology Department of Clinical Science Intervention and Technology Karolinska Institutet Huddinge Sweden

Citace poskytuje Crossref.org