This project is focused on research and development of new methods and tools for acquisition, computational analysis and efficient coding of plenoptic medical imaging content, specifically related to melanoma. Melanoma is known as the deadliest form of all types of skin cancer and, given its potential risk, accurate early detection of suspicious lesions is highly critical in what concerns to prevention strategies and treatment efficiency. Therefore, the project aims to advance the stateofart of technology and computational methods used in first diagnosis, which currently mainly relies on dermoscopy, a noninvase technique based on the inspection of RGB images. Although the efficiency of digital signal processing techniques has been increasing, and so improving the quality of diagnosis, the limited information provided by 2D RGB images is currently regarded as a significant constraint to achieve higher rates of concordance between expert observers, when assessing results of computational methods.
The project takes advantage of the most recent advances in imaging technology by relying upon melanoma plenoptic high resolution images, characterized by containing both spatial RGB intensity and directional information of light rays, captured by a light field (or plenoptic, or holoscopic) camera. Since such a rich visual content of melanoma light field does not exist yet, in any publicly available database, its acquisition and online access worldwide represents, on its own, a novel and useful outcome of the project. The increased amount of visual information contained in plenoptic high resolution images provide a unique data source for the project to carry out cuttingedge research on extraction of new features, taking advantage of a whole new range of possibilities, like image refocusing at difference planes, 3D manipulation within few microns of depth range, multiview processing and surface depth analysis.
As described, the project also aims to build the world first publicly available database of melanoma plenoptic images, for which efficient lossless compression algorithms will be investigated. Such research activity is focused on reducing the huge amount of raw data that is necessary to represent light fields beyond current standard codecs without losing any original information. The specific characteristics of plenoptic images, namely the regular patterns due to microlens array used by the camera and data redundancy will be exploited to devise more efficient compression methods, which will reduce the amount of storage capacity and bandwidth to access the database.
|Start Date: 01-04-2016|
|End Date: 01-12-2018|
|Team: Sérgio Manuel Maciel de Faria, Pedro Antonio Amado Assunção, Rui Manuel Fonseca Pinto, Luís Miguel de Oliveira Pegado de Noronha e Távora, João Miguel Pereira da Silva Santos, Pedro Miguel Marques Pereira, Jose Nunes dos Santos Filipe, Rui Miguel Leonel Lourenco|
|Groups: Multimedia Signal Processing – Lr|
|Partners: Centro Hospitalar de Leiria, Instituto Politécnico de Leiria|
|Local Coordinator: Sérgio Manuel Maciel de Faria|