Development of Image Processing Algorithm for Skin OCT Imaging

With the development of medical science, optical techniques are of particular importance in the medical field, and offer the therapeutic and diagnostic potentials. At present, the diagnosis of skin tumor often adopt skin biopsy technique in which a skin lesion is removed and sent to a pathologist to render a microscopic diagnosis. There are a lot of clinical experts are committed to the development of non-invasive methods such as ultrasound, dermoscopy, optical coherence tomography (OCT), confocal microscopy, photodynamic diagnosis technology, and so on. Especially OCT imaging technology can provide micrometer-resolution, high sensitivity and non-contact depth resolved cross-sectional images of biological samples, and is suitable for dermatology. However, hairs tended to create underlying shadow artifacts, and in sebaceous areas the many hair follicles and sebaceous glands tended to obscure normal layering. Also in skin cancer the cancer cell islands can mimic a hair follicle and make the ability using OCT to identify the location of skin cancer is greatly reduced. Therefore, how to combine the knowledge of image processing to analyze OCT images have become the research topics of optical image. In this project, we plan to develop a comprehensive, automatic OCT image processing computer-aid diagnosis (CAD) platform. Combined removable optical tomography system with multi-imaging technology developed by other sub-projects and verified by clinicians, to construct an appropriate overall quantitative tool that provides dermatologists to diagnose skin tumor and improve treatment at the earliest possible time. This tool can not only effectively track the recovery situation of patient after surgery but achieve the applied goal integrate clinical and engineering. Therefore, we will develop suitable image enhancement algorithms, automated skin layer segmentation procedure, and initial vessel segmentation algorithm for OCT images in first year. For the second year, we will optimize our OCT image enhancement and vessel segmentation algorithms to increase the adaptability and accuracy for different types of skin cancer. In addition, we will undertake the exploitation of three-dimensional OCT image reconstruction techniques to build 3D visualization module that can enhance the visibility and correlations between vessels and tumor. We will integrate fruitful results of first and second year and focus on the regions of angiogenesis to develop OCT image information analysis and statistical algorithms that can obtain quantitative information to establish multiple quantitative indexes as a basis for clinical evaluation in third year. This system we proposed can not only assist skin science, imaging science, and researchers in biomedical optics in acquiring more information but also establish the correlations between multiple types of information. The research results can facilitate the search for the unknown clinical questions.

Project ID:PB10308-2692
External Project ID:MOST103-2221-E182-038