HIQ
Software for classification of pixels in a digital image of a biological sample
This technology has been patented since 2021 under the number EP3821396, extended in Australia, Canada, China, Japan and United-State and co-owned University of Caen, France, National Institute of Health and Medical Research, France, François Baclesse Regional Center, France and University of Vilnius, Lithuania.
Description:
A software for the detection of colorations in histology which allows a fully automated classification of biomarkers expression and guides pathologists in their decision. Any type of digital image can be used and manual calibration can be done by the user for a greater transparency.
How does it works:
The software is based on two mathematical algorithms:
- Principal Component Analysis (PCA) for changing color space from RGB and in particular to OHTA space better adapted to histology;
- Gaussian Mixture Method (GMM) to search in the histogram colors two gaussians separating noise from areas of interest and two gaussians separating stained pixels from non-stained pixels.
A binary mask is applied on the image to highlight stained pixels, then the image is superimposed with a hexagonal macro-pixel, allowing a better heterogeneity and a better detection of stained areas.
Applications:
It applies to many clinical situations in which immunomarkers need to be characterized and quantified and has already been tested for pathology such as ovarian cancer and breast cancer.
It also can be used to build learning databases for Deep Learning networks.
Advantages:
Eye-only | Our software |
| Analysing and quantifying stained cells is a laborious and time-consuming task for pathologists. It also remains subjective due to interobserver and intraobserver variations. | It allows an objectif analysis which leads to a better reproducibility of the results. |
Machine-learning technologies | Our software |
| The results obtained often can't be explained due to the "black-box effects". Bias can also be introduced if the training databases are not representatives. | False-positive and false-negative can be avoided because the user can choose its own settings. |
-
expand_more library_books References (2)
- Marylène Lejeune, Benoît Plancoulaine, Nicolas Elie, Ramon Bosch, Laia Fontoura, Izar de Villasante, Anna Korzyńska, Andrea Gras Navarro, Esther Sauras Colón & Carlos López , How the variability between computer-assisted analysis procedures evaluating immune markers can influence patients’ outcome prediction
- Nicolas Elie, Florence Giffard, Cécile Blanc-Fournier, Pierre-Marie Morice, Pierre-Emmanuel Brachet, Soizic Dutoit, Benoît Plancoulaine, Laurent Poulain , Impact of automated methods for quantitative evaluation of immunostaining: Towards digital pathology
-
expand_more cloud_download Supporting documents (2)Datasheet_englishPlease find further information about this technology in the following datasheet.HIQ datasheet.pdf (431 KB)Fiche technique_françaisVous trouverez plus d'informations sur cette technologie dans la fiche technique ci-contre.Datasheet HIQ fr.pdf (586 KB)Additional files may be available once you've completed the transaction for this product. If you've already done so, please log into your account and visit My account / Downloads section to view them.