• HOME
  • News
  • Blood cell classification support system(NGX-029A)

Blood cell classification support system(NGX-029A)

  • 2022/04/13

This system supports diagnosis for myelodysplastic syndrome(MDS).

Challenges in Clinical Practice

MDS (myelodysplastic syndrome) is a disease in which hematopoietic stem cells in the bone marrow become abnormal and stop producing normal blood cells.

The decrease in normal blood cells causes symptoms such as anemia, bleeding tendency, and fever associated with infection. Observed characteristics of the disease include “ineffective erythropoiesis ,” in which blood cells are destroyed in the process of becoming mature cells, and “dysplasia,” in which the formed blood cells are abnormally shaped. In addition, some patients may develop MDS or even  “acute leukemia”.

In the diagnosis of MDS, a statistical report such as blood cell counts is used after clinical technicians observe bone marrow smear images, classify blood cells according to their morphological characteristics..

However, technicians need to observe approximately 500 smear images per patient and classify more than 30 cell types, which is very burdensome, and the  classification accuracy depends on their experience.

Our Technology

In order to speed up the classification and counting of blood cells by technicians, NextGeM developed an

image segmentation AI that automatically detects, classifies, and counts blood cells based on the input images taken from bone marrow smears. (*This program is not intended for diagnosis, treatment, or prevention of diseases.)

We trained the image segmentation AI to learn the morphological characteristics of blood cells, using blood cells labeled by physicians and technicians at Kyoto University Hospital as training data (approximately 100 cases and tens of thousands of images), .

Since this SaaS system is linked to the image segmentation AI model deployed in the cloud, it enables remote consultation with physicians and efficient information sharing between physicians and technicians who are separated even in the same hospital.

A function that allows technicians to correct the output results by the image segmentation AI is also under development.