Umeå universitets logga

Every year approximately 2800 Swedes are diagnosed with malignant melanoma, the form of cancer that is most rapidly increasing in incidence in the Western world. The earlier we can identify and diagnose a malignant melanoma, the better is the prognosis. In Sweden, 155 000 benign naevi, harmless skin tumours or moles, are surgically excised each year, many of them because melanoma cannot be dismissed by non-invasive methods. The excisions result in substantial medical costs and cause unrest and suffering of the individual patient. For untrained physicians, it is often difficult to make an accurate diagnosis of melanoma, thus a tool that could help to strengthen the diagnosis of suspected melanomas would be highly valuable. This thesis describes the development and assessment of a non-invasive method for early skin cancer detection. Using near infrared (NIR) and skin impedance spectroscopy, healthy and diseased skin of various subjects was examined to develop a new instrument for detecting malignant melanoma. Due to the complex nature of skin and the numerous variables involved, the spectroscopic data were analysed multivariately using Principal Component Analysis (PCA) and partial leas square discriminant analysis (PLS-DA). The reproducibility of the measurements was determined by calculating Scatter Values (SVs), and the significance of separations between overlapping groups in score plots was determined by calculating intra-model distances.

The studies indicate that combining skin impedance and NIR spectroscopy measurements adds value, therefore a new probe-head for simultaneous NIR and skin impedance measurements was introduced. Using both spectroscopic techniques it was possible to separate healthy skin at one body location from healthy skin at another location due to the differences in skin characteristics at various body locations. In addition, statistically significant differences between overlapping groups of both age and gender in score plots were detected. However, the differences in skin characteristics at different body locations had stronger effects on the measurements than both age and gender. Intake of coffee and alcohol prior to measurement did not significantly influence the outcome data. Measurements on dysplastic naevi were significantly separated in a score plot and the influence of diseased skin was stronger than that of body location. This was confirmed in a study where measurements were performed on 12 malignant melanomas, 19 dysplastic naevi and 19 benign naevi. The malignant melanomas were significantly separated from both dysplastic naevi and benign naevi.

Overall, the presented findings show that the instrument we have developed provides fast, reproducible measurements, capable of distinguishing malignant melanoma from dysplastic naevi and benign naevi non-invasively with 83% sensitivity and 95% specificity. Thus, the results are highly promising and the instrument appears to have high potential diagnostic utility.