How to Classify Blood Cells Using Neural Networks

Introduction

Machine Learning is widely used for image recognition. Here, we tried to perform blood cell classification using Convolutional Neural Networks.

There is a dataset of blood cells that can be used for training and testing [2, 3].

According to the paper [1], the model that performed the best is the Regionnal Convolutional Neural Network (R-CNN).

We tried to reproduce their results.

The dataset

The dataset is a set of images of white blood cells and platelets sorted in 8 classes: immature granulocytes (ig), monocytes, basophils, neutrophils, eosinophils, erythroblasts and platelets.

The dataset contains 17092 images.

The model

Firstly, we tried to use a classic Convolutional Neural Network (CNN).

References

[1] White blood cells detection and classification based on regional convolutional neural networks

[2] A dataset of microscopic peripheral blood cell images for development of automatic recognition systems

[3] The precedent dataset, download page