Automatic analytical software for interpreting skin properties

点击翻译按钮获取中文摘要

Medical evaluation and diagnosis of skin diseases primarily rely on visual inspection for morphologic features such as colour, shape, border, configuration, distribution, elevation, and texture. The overall clinical approach is subjective and qualitative. Over the past 20 years, there has been rapid development in the medical imaging hardware. However, these new devices still require human experts for interpretation. There is a critical gap between rapidly advancing hardware development and the development of automatic analytical software. In this application, I propose to bridge the gap by working on image analysis programs for the new imaging modalities that could be used for early detection of melanoma, the most dangerous form of skin cancer. In particular, I will work on the following three areas: (1) I will develop new methods to analyse images of internal skin structures acquired by dermoscopy, a hand-held device often used by dermatologists to examine suspicious moles. (2) I will build a set of specialized programming tools for a new optical device that uses light to probe the skin's internal structures to detect abnormality and malignancy. (3) I will also create a new 3-D program to match moles between photographs in order to facilitate physicians who practice total-body photography to follow high-risk melanoma patients. When the proposal is completed, we will find new ways to analyse complex images ranging from 2-D to 4-D, and from greyscale to colour. The new knowledge will benefit the imaging community because the new methodologies can be applied to other image processing and analysis applications. In addition, our new computer programs, especially designed for melanoma detection as set out by the proposal, will move us one step closer to achieving a fully automated skin lesion diagnostic system, which will ultimately help diagnose melanoma in the early stages, thereby reducing un-necessary skin biopsies and decreasing mortality rate.