Quantification of structures of skin lesion and calibration of dermoscopy images for automated melanoma diagnosis

为了使我们基于互联网的黑色素瘤筛查系统适合实际使用，我主要研究了以下两个主题。(1)皮肤镜结构的定量分析(2)皮肤镜图像颜色自动校准技术的发展。就这两个主题而言，我已成功地实现了建议中规定的目标。(1)成功建立了一个识别系统，在识别ABCD规则和7点检查表中定义的共15个皮肤镜结构方面与皮肤科专家无统计学差异。(2)在没有任何特殊设备的情况下，对皮肤镜图像在亮度、色调和饱和度上实现了等效的色彩校准性能。

I had investigated mainly following two themes for making our Internet-based melanoma screening system fit for practical use. (1) Quantification of dermoscopic structures (2) Development of automated color calibration for dermoscopy images. For both themes, I had successfully achieved the objectives set out in the proposal. (1) Succeeded to build a recognition system that has statistically no difference with expert dermatologists in recognizing a total of 15 dermoscopic structures defined in ABCD rule and 7-point checklist. (2) Achieved equivalent color calibration performance for dermoscopy images in luminance, hue and saturation without any special devices.