Application of Machine Learning Algorithm in Identification of Anaemia Diseases

Abstract

Anaemia is the reason for lot of serious health-related issues, and if it is detected at an early stage accurately, it can help to avoid problems like fatigue, pregnancy complications, heart problems, and life-threatening complications. In this study, five supervised machine learning (ML) algorithms for detecting anaemia based on red blood cell (RBC) parameters such as haemoglobin, haematocrit or packed cell volume (PCV), RBC count, mean corpuscular volume (MCV), mean corpuscular haemoglobin (MCH) and mean corpuscular haemoglobin concentration (MCHC) were developed. The ML algorithms were applied on actual patient anaemia data, collected from a clinical laboratory, of a sample size of 2000 data samples. Multiclass classification of these data into five different conditions of anaemia, namely beta thalassemia trait (BTT), dimorphic anaemia (DA), and macrocytic blood picture (MBP), microcytic hypochromic anaemia (MHA), including normocytic normochromic blood picture (NNBP) indicating no anaemia, was performed by the five ML algorithms, and the best algorithm for accurately detecting anaemia was identified. The result indicated that the decision tree and random forest algorithms were superior to other algorithms in terms of accuracy, sensitivity, and specificity of the identification process.