THALASSEMIA CARRIER IDENTIFICATION FROM BLOOD MICROSCOPY IMAGES: A COMPARATIVE ANALYSIS OF ML AND DL APPROACHES

Abstract

Thalassemia is one of the most prevalent genetic disorders globally, especially in regions like Asia and among individuals of African descent. In Pakistan, the carrier rate for thalassemia is between 5% and 7%, affecting over 10 million people, with approximately 5,000 children born annually with β-thalassemia major (β-TM). Early detection of carriers is vital for managing and preventing severe cases. This study introduces a machine learning approach to detect thalassemia carriers through blood smear image analysis. We preprocessed the images to extract features such as color, texture, and shape from a dataset of 7,108 blood images, representing nine cell types related to thalassemia. Various machine learning and deep learning models were applied to classify the images as thalassemic or non-thalassemic. Among the machine learning models, Random Forest achieved the highest accuracy at 91.1%, while MobileNetV2 led among deep learning models with a 90% accuracy. These promising results suggest potential for real-world application in Thalassemia screening programs, contributing to improved diagnostic methods and healthcare outcomes for affected populations.