Advancing Anemia Diagnosis: Harnessing Machine Learning Methods for Accurate Detection

Anemia, a condition characterized by insufficient healthy red blood cells to transport oxygen adequately, demands timely detection and treatment to prevent complications. Machine learning methods offer promising avenues for diagnosing anemia, especially in settings with limited medical expertise and resources. This study employed various algorithms, including k-nearest neighbor, support vector machines, decision trees, adaptive boosting, and random forest, to diagnose anemia and its specific types. The dataset encompassed laboratory data from 5,553 anemia patients and 9,747 healthy individuals, with subtypes including iron deficiency-related anemia, hemoglobin insufficiency, B12 deficiency, and folate deficiencies. Using 24 features like B12 levels, blood cell counts, iron levels, and gender, the study aimed to identify the anemia presence and recommend treatment. Notably, the decision tree algorithm emerged highly effective, surpassing other methods with a 96.49% accuracy rate. The study achieved a remarkable 100% accuracy in anemia diagnosis by optimizing hyperparameters through grid search. This outcome holds significant clinical implications, as misdiagnosing anemia can lead to severe complications like hemochromatosis, potentially resulting in organ damage or failure.