CBC and DBC Counter Using Image Processing

Clients

Techstack

Purpose

Counting blood cells by hand under a microscope is slow, tiring, and inconsistent from one technician to the next — and a diagnosis can hinge on getting it right. A research team wanted to prove that a machine could read a microscope image of a blood sample and produce complete and differential counts on its own, fast and consistent enough to help a lab.

Description

We built a device that captures a blood-sample image and uses a deep-learning vision model to identify and count red cells, white cells, and platelets, with image processing cleaning up and separating the cells before they are classified. The results appear on a simple on-device screen, and — because clinics are not always online — the device can text them out over the cellular network with no internet connection at all. It is exactly the kind of high-stakes medical computer-vision problem we take on and ship.