Identification and patient blood subtypes

Diagnosing blood diseases often involves the identification and characterization of patient blood samples. Mathematical neural-type methods can be very useful in the automated recognition of blood cell subtypes. By having this data available, mass processing of these allow the implementation of an automatic detection model and classification of cells such as eosinophils, lymphocytes, monocytes and neutrophils. The goal of the study "Convolutional Neural Network and decision support in medical imaging: case study of the recognition of blood cell subtypes", published in the CEUR Workshop Proceedings, ISSN: 1613-0073, is to use the learning Convolutional Neural Network (CNN) type deep machine for the recognition of blood cell type images and to make them capable of classifying them such as eosinophils, lymphocytes, monocytes or neutrophils. Accuracy of classification on all learning data is 97.39% and the validation accuracy is 97.77%. Failure of image detection is very low.

Architecture of the proposed CNN classifier. The input represents a 2D image, followed by convolution layers and max pooling layers to compute n sets of 32 then 64 classified feature maps with a fully connected network
CNN Classifier is a great method for identification of blood cells.