With an IHC or IF slide, the Immuno Object by Learning enables identification of objects featuring specific characteristics so they can be placed into two categories: the objects possessing these characteristics and those that do not. Based on machine learning techniques, this algorithm is able to identify and quantify categories of objects that are very different from each other..
It is particularly useful to quantify cellular staining in both IHC and IF :
- Nuclear: Ki67, RO, RP
- Cytoplasmic: CD68, CD163
- Membranes: CD3, CD8, CD20
- IF multiplexing: DAPI/FITC/TRITC/CY5
The protocol :
The different stages of image analysis using the IOBL algorithm are as follows :
- Creation of a learning model :
The RFT tool enables a learning base to be created that enables cells presenting certain characteristics to be identified. The user then stains the different objects using a brush so that the algorithm can automatically determine the classificatory characteristics.
- Choice of sampling strategy
when analysis results are being calculated, the region of analysis is cut into sections. According to predetermined criteria, it possible to analyze some or all of the sections on the basis of statistical sampling strategies.
- Choice of the type of result and post-processing
the user can choose to count the objects or to generate a result describing each object. In this case, morphomathematical operations allow objects to be selected according to certain criteria.
- Definition of staining positivity thresholds
the user can sort the segmented objects into two categories, positive or negative, or four categories: 0, 1+, 2+ and 3+.
Results are expressed in the number and percentage of objects for each of the four categories. They are compared with the total number of objects present in the region of analysis. Analyzed objects can also be individualized and described on a one-by-one basis.