Dermoscopic Image Quality
The Future Projects that LIVE is targeting are problems of great significance for
which there currently exist no adequate solution. Medical image quality is an
important problem that has unfortunately received relatively little attention, owing to
its cross-disciplinary demands, and because of a strong focus by funding agencies on
imaging device improvement, as opposed to demonstrating the perceptual efficacy
of these devices, or optimizing medical imaging devices and algorithms for human
(physician) viewing and interpretation. While we have an exceptional grounding in
perceptual image quality modeling and prediction, the next problem of applying
these models to study the effects of perceptual image quality on visual task (such as
finding a lesion in a mammogram, or in a dermoscopic image) remains wide-open
and largely unstudied.
LIVE has a very strong research history in medical imaging, e.g., in digital mammography, our current greatest interest is in Dermoscopy in the Wild. Malignant melanoma is occurring at a rapidly increasing pace throughout the globe, for a variety of factors related to human activity, climate change, population growth, and lack of education and especially access to healthcare. We are greatly interested in helping to develop mobile imaging solutions for dermoscopy towards advancing melanoma and other skin cancer detection “in the wild,” in remote, rural, and poverty-stricken locations throughout the world. While mobile dermoscopy is not a new idea, the concept of using powerful perceptual picture quality tools to inform the local (possibly non-physician) dermoscopist regarding the quality of the dermoscopic pictures taken, to improve the images obtained or to cue that a different image be acquired, and to better assist the remote dermoscopist towards making distant diagnoses by informing the process of the nature of the perceptual picture quality and how it might affect the inspection process. It is our view that millions of lives may be affected by the development of such “quality-aware” dermoscopic imaging tools, extending hugely successful Television and Cinema techniques to a life-saving enterprise of global consequence. We have just begin work in this direct with interested colleagues and a former post- doc in LIVE:
F. Xie and A.C. Bovik, “Automatic segmentation of dermoscopy images using self- generating neural networks seeded by genetic algorithm,” Pattern Recognition, vol. 46, no. 3, pp. 1012-1019, March 2013.
F. Xie, Y. Lu, A.C. Bovik, Z. Jiang and R. Meng, “Application-driven no reference quality assessment for dermoscopy images with multiple distortions,” IEEE Transactions on Biomedical Engineering, vol. 63, no. 6, pp. 1248-1256, June 2016.
F. Xie, H. Fan, Y. Li, Z. Jiang, R. Meng and A.C. Bovik, “Melanoma classification on dermoscopy images using a neural network ensemble model,” IEEE Transactions on Medical Imaging, vol. 36, no. 3, pp. 849-858, March 2017.