Detecting Pediatric Acute Lymphoblastic Leukemia using Computer Vision

Local Chapter Liverpool, England Chapter

Coordinated by,

Status: Completed

Project Duration: 11 Mar 2023 - 15 Apr 2023

Open Source resources available from this project

Project background.

In ALL there is an accumulation in the bone marrow of immature lymphocyte precursor cells, called blast cells. Eventually, the production of normal blood cells is affected by this, resulting in a reduction in the number of red cells, normal white cells, and platelets in the blood.

ALL is the only form of leukemia that is more common in children than adults. It is the single most common form of pediatric cancer, accounting for about one-third of all cases in children. About 85% of cases of childhood leukemia are ALL and it occurs in about 400 children in the UK each year. ALL occurs mostly between the ages of about two and four years. Males are affected more often than females at all ages.

The problem.

The task of identifying immature leukemic blasts from normal cells under the microscope is challenging due to morphological similarity, this can make the diagnosis time consuming and costly.  

Project goals.

About 85% of cases of childhood leukaemia are ALL and it occurs in about 400 children in the UK each year. The goal of the project is to build a model which will make early identification of paediatric ALL quick and cost effective – this will allow early treatment to be sought and potentially save the lives of those who are afflicted. 

Project plan.

  • Week 1

    Week 1:
    Data collection / organisation

  • Week 2

    Week 2:
    Data cleaning / augmentation / engineering

  • Week 3

    Explore the images and any augmentations with analysis

  • Week 4

    Build and test a computer vision model

  • Week 5

    Develop an app for inference

  • Week 6


Learning outcomes.

Data collecting / organisation, Data cleaning / augmentation / engineering, Exploratory Analysis, Building a Computer Vision model, Develop and deploy an app.

Share project on: