Computer Aided Diagnosis of Melanoma Based on Confocal Laser Microscopy

  • Wanle Chi
Keywords: Microscopic Examination, Computer Aided, Feature Extraction, Image Recognition


Melanoma is a kind of malignant skin tumor which is often encountered in clinic. It is also one of the fastest
growing cancers in the world. Usually, dermatologists use the methods of naked eye observation and
histopathological biopsy to screen and diagnose melanoma in early stage. Blind biopsies often cause financial
stress and unnecessary physical trauma. Therefore, non-invasive melanoma automatic diagnosis technology has
become an urgent problem in the medical field. Based on the above background, the purpose of this paper is to
study the computer-aided diagnosis algorithm of melanoma based on the image of laser confocal microscope. In
this paper, the texture features of laser confocal scanning microscope images of melanoma and common benign
nevus are extracted by wavelet analysis. Based on the standard deviation, energy and entropy characteristic
parameters of wavelet coefficients, the image is automatically classified by classification and regression tree
algorithm. The experimental results show that the framework can achieve 3.4% improvement of AUC index
compared with the baseline represented by the common single model. Finally, an integrated recognition model
based on AUC weighted average is proposed. The AUC of the integrated model is 7.7% higher than that of
baseline, which has some advantages over the current mainstream algorithm model. The experimental results
verify the effectiveness of the proposed method and model. Therefore, the research results of this paper have a
good reference value for the research of melanoma detection technology in dermoscopic images.