Moichor

ML-enhanced automatic diagnosis system for biomedical blood tests

MLaaS, Web

Objectives and challenges

Moichor is a Canadian animal diagnostics service that provides among other things blood testing enhanced with machine learning. Representatives of Moichor entrusted SOLVVE to find a solution that would help to provide the most accurate blood test results, which is possible due to the elimination of the subjectivity from manually performed tests.

Solutions

SOLVVE used knowledge databases containing a big amount of high-resolution images of cells to help the doctors diagnose the conditions based on the blood tests. Our experts used the transfer learning technique in conjunction with the fine-tuning of the model’s hyperparameters to train the algorithm to detect specific kinds of blood cells. Object detection combined with instance segmentation and classification allows this technology to deliver extremely precise results.

Features

Resistance to changes in the color map.
Data augmentation.
Possibility to add new classes as the diagnosіs count grows.
Automatic updates of image database.
Automatic re-training of the system based on the database updates to boost results accuracy.