SIR, Dermoscopy is a powerful instrument especially designed for the diagnosis of pigmented skin lesions (PSLs). However, it is certainly subjective and therefore associated with low reproducibility and potential errors in the diagnosis of PSLs. Computer analyses of PSLs could overcome the subjectivity of dermoscopy. Several groups have already developed automated analysis software in order to solve these problems and reported the high sensitivity and specificity of their algorithms. Nevertheless, we think that there are still the following problems in software-based analysis reported in the above studies. 1 The results of these studies are not comparable, because different images are used in each study and there is no image standardization. Although the authors sometimes defined their images as equivocal naevi or dysplastic naevi, the meanings of the terms are different among dermatologists. 2 The number of dermoscopic images subjected to digital analysis in the above reports was not sufficient for proper statistical analysis, and the images were usually collected in one or a few institutions. The greater the number of images used in a study, the more accurate and the more reliable the study certainly becomes. 3 The use of personal computer-based analysis programs, rather than server-based programs, to calculate parameters for discrimination is a problem. Although such applications are useful on a single desktop computer, they are not suitable when planned for use by many users because installation and upgrade of such programs are often troublesome. To solve these problems, a website containing a Java program on a Linux server system has been established at our university. The complete URL of the site is http:// dermoscopy.soft.ics.keio.ac.jp/index_eng.html. The site is designed so that any visitor can upload the digital dermoscopic images and register the clinical and pathological data. In addition to allowing uploading of the images, our website provides discrimination analysis between melanoma and Clark naevi. After the server machine accepts an image, the program automatically starts to extract the PSL image from the surrounding skin, and calculates several parameters from both the entire lesion and the 10% periphery of the lesion. The lesion is evaluated as a melanoma or Clark naevus using the above parameters and the result (melanoma or Clark naevus) is sent back to the client. A sensitivity of 87Æ0% and specificity of 93Æ1% have been calculated based on a preliminary study of 77 cases of melanoma (including 23 cases of melanoma in situ) and 188 cases of Clark naevi. The use of this internet-based program offers certain advantages. 1 Our system could easily create a large library of images. The uploaded images are stored on the server and can be offered for viewing to other dermatologists. The latter could use the library images and select the images to use for their analyses. The use of the same images could allow comparison of the analytical power of different digital discrimination analyses. 2 The client may not necessarily need to use a personal computer. A portable digital assistant (PDA) or a mobile phone, with a digital camera and polarized filters, could be used in the future. 3 The web site is available 24 h to anyone with access to the internet. 4 Following the release of the update version, the latest version of the program is always available all the time for all users. Users do not have to worry about problems related to the operating system, installation or updating. We have already noticed that there were some drawbacks in our system. At this point, we do not have a program that could be applied to other PSLs except melanoma and Clark naevus. Our program cannot be applied to mucous or acral lesions. As described earlier, digital extraction of the lesion from the surrounding skin is the most difficult step. Lesion acquisition could fail when the image is not completely within the frame or the lesion contains some hairs. However, we continue to improve our program to overcome these weak points and hope eventually to create an auto-screening system for all types of PSLs in the near future.