Kittler H, Seltenheim M, Pehamberger H, Wolff K, Binder M
Department of Dermatology, University of Vienna Medical School, Austria.
Melanoma Res. 1998 Jun;8(3):255-60. doi: 10.1097/00008390-199806000-00008.
Epiluminescence microscopy (ELM) is a useful method for improved diagnostic accuracy in early cutaneous melanoma. Conventional photographs of ELM images are commonly used for clinical research and documentation. Electronic images have advantages compared with photographs and are essential for medical informatics, computerized learning and telemedicine. Compression of electronic images allows a reduction in volume of data, but significant image deterioration may occur at high compression rates. We sought to study the diagnostic informativeness of compressed digital ELM images compared with conventional photographs. Fifty photographs of pigmented skin lesions, including 23 melanomas, were presented to eight dermatologists as photographic slides and as digital images with 30:1 Joint Photographic Experts Group (JPEG) compression. The diagnostic performance of the media and the readers was described in terms of sensitivity, specificity and areas under receiver operating characteristic curves (AUC). Agreement between the readings of the two types of media regarding the presence or absence of ELM criteria was assessed using kappa (kappa) statistics. The mean AUC was 0.81 (95% confidence interval [CI] = 0.73-0.90) for slides and 0.81 (95% CI = 0.72-0.90; P = 0.89) for digital images. Agreement between the readings of the two types of media regarding the presence or absence of ELM criteria ranged from kappa = 0.55 (95% CI = 0.22-0.88) for grey-blue area to kappa = 0.89 (95% CI = 0.74-1.00) for radial streaming. In conclusion, digital ELM images with 30:1 JPEG compression appear to be as informative as photographic slides when used to differentiate between melanoma and non-melanoma.
表皮透光显微镜检查(ELM)是一种有助于提高早期皮肤黑色素瘤诊断准确性的有用方法。ELM图像的传统照片通常用于临床研究和记录。与照片相比,电子图像具有优势,并且对于医学信息学、计算机化学习和远程医疗至关重要。电子图像的压缩可减少数据量,但在高压缩率下可能会出现明显的图像劣化。我们试图研究与传统照片相比,压缩数字ELM图像的诊断信息量。向八位皮肤科医生展示了50张色素沉着性皮肤病变的照片,其中包括23例黑色素瘤,分别以幻灯片形式以及具有30:1联合图像专家组(JPEG)压缩的数字图像形式呈现。根据敏感性、特异性和受试者操作特征曲线下面积(AUC)来描述介质和读者的诊断性能。使用kappa(κ)统计量评估两种类型介质在ELM标准存在与否方面的读数之间的一致性。幻灯片的平均AUC为0.81(95%置信区间[CI]=0.73 - 0.90),数字图像的平均AUC为0.81(95%CI = 0.72 - 0.90;P = 0.89)。两种类型介质在ELM标准存在与否方面的读数之间的一致性范围从蓝灰色区域的κ = 0.55(95%CI = 0.22 - 0.88)到放射状条纹的κ = 0.89(95%CI = 0.74 - 1.00)。总之,当用于区分黑色素瘤和非黑色素瘤时,具有30:1 JPEG压缩的数字ELM图像似乎与照片幻灯片具有相同的信息量。