Date of Award
Doctor of Philosophy
Angie Izard, Jun Zhang, Gary L Williams, Chao Zhu, Ethan Munson, Zeyun Yu
3D modeling, CNN segmentation, Human anatomy labeling system, Mobile wound assessment, Projective texture mapping, Wound imaging
The prevalence of camera-enabled mobile phones have made mobile wound assessment a viable treatment option for millions of previously difficult to reach patients. We have designed a complete mobile wound assessment platform to ameliorate the many challenges related to chronic wound care. Chronic wounds and infections are the most severe, costly and fatal types of wounds, placing them at the center of mobile wound assessment. Wound physicians assess thousands of single-view wound images from all over the world, and it may be difficult to determine the location of the wound on the body, for example, if the wound is taken at close range. In our solution, end-users capture an image of the wound by taking a picture with their mobile camera. The wound image is segmented and classified using modern convolution neural networks, and is stored securely in the cloud for remote tracking. We use an interactive semi-automated approach to allow users to specify the location of the wound on the body. To accomplish this we have created, to the best our knowledge, the first 3D human surface anatomy labeling system, based off the current NYU and Anatomy Mapper labeling systems. To interactively view wounds in 3D, we have presented an efficient projective texture mapping algorithm for texturing wounds onto a 3D human anatomy model. In so doing, we have demonstrated an approach to 3D wound reconstruction that works even for a single wound image.
Williamson, Victor, "Mobile Wound Assessment and 3D Modeling from a Single Image" (2020). Theses and Dissertations. 2624.