Optimizing Illumination Color in Skin Lesion Classification
Clayton Shafaer
cms156@duke.edu
| Paper PDF |
|
Correct skin lesion classification can be the difference between life and death for patients with early stage skin cancer. While the task is typically done by a trained professional, automated skin lesion classification is an area of ongoing research in the machine learning community. The goal of many of these approaches is to provide increased accessibility to these types of healthcare services. While visiting a dermatologist's offices can have many barriers associated with it, such as health insurance and time off work, a machine learning application that classifies skin lesions is much more accessible. In this project, I will explore physical optimizations that we can apply to skin lesion classification, with the goal of increasing the accuracy of mobile health skin lesion imaging systems (such as smartphone apps). Specifically, I will be exploring the optimal illumination color for this task, arriving at a result that can be applied to skin lesion images through a colored flash or a variable color LED array.
|
|
|
| Paper: |
Code and Data:
|