ARTIFICIAL INTELLIGENCE
To the new digital assistants
Pixience is one of the main French stakeholders in the development of artificial intelligence in the field of healthcare. We intervene to design intelligent systems, in all disciplines producing images of living organisms and whose analysis requires a high level of expertise.
Artificial Intelligence for Dermatology
The Artificial Intelligence designed by Pixience was trained on a large database of images of the C-Cube system qualified by a network of expert dermatologists to distinguish melanomas from benign nevus.
But this is only the first step. The technology implemented will make it possible to go much further, by providing assistance on other complex skin pathologies during consultations.
Highlight your expertise
Artificial Intelligence is a tool that will not replace the eye of the specialist. But these new digital decision support assistants will provide support to practitioners, to refine and confirm their diagnoses.
If you want to join our community of dermatologists and participate in the development of our Artificial Intelligence, contact us!
Constantly advancing technology
AI uses deep learning techniques, an approach that is still young in the health sector, but with considerable potential. It aims to design computer systems inspired by the biological functioning of the human brain: convolutional neural networks (CNNs).
Their goal is to learn complex processes independently. These neural networks, adequately trained on annotated image bases, will then be able to give an automatic analysis, ideally comparable to that of an expert, of new images.
A unique technological lead of the C-Cube system
Deep Learning model classification is notoriously well suited to problems with high variability, so it’s a perfect approach for pigment spot analysis. However, conventional dermoscopy devices add to this intrinsic variability, additional variations induced by zoom or lighting effects, which deteriorate the performance of algorithms.
Indeed, the algorithm learns through the analysis of the characteristics of the lesion: shape, texture, regularity, colors, brightness, each parameter is essential for a correct diagnosis. The quality of the digital image of the dermatological lesion is therefore decisive not to induce ultimately a diagnostic error.
Thanks to the patented 2D & 3D technology of the C-Cube dermoscope, whose metrics and colors are calibrated (TrueColor Technology), we have achieved some of the highest performance in research in the field, while reducing the complexity of learning models.
Thus, during the validation tests of our neural network, the Pixience classifier correctly detected more than 90% of the skin cancers observed with the C-Cube dermoscope.
