Ugly duckling feature for melanoma detection : a validation study – DIAMELA
An ugly duckling mole: I'm interested
Most of the melanomas are recognizable from their visual characteristics, especially significant when they differ from those of the other nevi of the individual observed (ugly duckling). Rarely described in the literature, this idea is tested and evaluated for impact.
Validation and computer implementation of the ugly duckling concept
The first phase of this project is to experimentally validate the Ugly duckling concept by analyzing the consistency of the expertise provided by well known international dermatologists for the nevus of one hundred patients. An automatic diagnosis system based on the ugly duckling concept will be built and validated during the second phase. It will be, in a third phase, integrated with an existing system of diagnosis of melanoma (resulting from the previous ANR project SKINAN) for an assessment study in clinical routine. A remote system for dermoscopic image acquisition will be built for the occasion.
The project essentially comprises three phases: 1. gathering expertise relevant to the problem and experimental validation of the ugly duckling concept by the analysis of the correlation between dermatologists, 2- identifying common characteristics of nevi for each patient and highlighting the original look of some of them, by image analysis methods combined with high dimensional data methods based on learning by example, 3- integrating the ugly Duckling concept in an pre-existing system for the automatic melanoma diagnosis and evaluating its contribution. The methodological problems encountered in this study revolve around the search for consensus among experts, big data analysis and learning of structured data.
Two major results: The Ugly Duckling feature is an important element of melanoma diagnosis. It reduces significantly the number of tumor removal without impacting the quality of diagnosis – An original method of data mining applied to the classification of objects inferred from examples has been designed. Many applications are possible.
Outlook: Establish, notably through examples, the elements for detecting Ugly duckling moles that can be useful for postgraduate education and training for self-diagnosis.
The construction of whole body cabins for the overall analysis of moles and also track their evolution is strongly encouraged. The availability to the public of a connected dermoscope could be very helpful particularly in a context of teledermatology.
12 articles and proceedings including the proof of the Ugly Duckling concept: Journal of Investigative Dermatology, October 2013, Vol. 133, No. 10, p. 2355-61, 201 + 2 articles submitted: impact of the Ugly Duckling concept for diagnosis, description of the method. One invention Declaration
Project coordination
Bernard FERTIL (Laboratoire des sciences de l'information et des systèmes) – bernard.fertil@lsis.org
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
Partner
LSIS Laboratoire des sciences de l'information et des systèmes
CNRS - FRE3312 Centre national de la recherche Scientifique
APHM service de dermatologie de la Timone, Marseille
SATT PACA Corse Société d'Accélération du Transfert de Technologies
Help of the ANR 329,773 euros
Beginning and duration of the scientific project:
January 2013
- 24 Months
