A comprehensive review of machine learning approaches for detecting anemia and abnormal red blood cells using RBC indices and medical imaging

Anemia affects more than one billion people globally. It is caused by low hemoglobin levels resulting from iron or vitamin deficiencies, blood loss or genetic conditions like sickle cell anemia and thalassemia. Anemia diagnosis is challenging due to costly blood tests, lack of medical personnel and limited infrastructure especially in rural areas. Machine learning models can analyze the records of complete blood count and images of blood smears, eye conjunctiva, lip mucosa, fingernails, and palms for anemia detection, and therefore eliminate the need for expert healthcare professionals. This study reviews artificial intelligence techniques to diagnose three types of anemia: iron deficiency, sickle cell disease and thalassemia. For each type of anemia, we have compared machine learning, deep learning, optimization algorithms, hybrid algorithms and other techniques in terms of accuracy, key features, research gaps and future scope. We have also assessed the dataset sizes and their sources used in these studies to explore any correlation with models accuracy. Artificial intelligence can be integrated into real-time anemia diagnosis as it offers non-invasive, quick, affordable and accurate solution with explainable AI techniques to enhance physicians trust and usability particularly in remote areas. However, traditional invasive tests, such as the complete blood count and peripheral blood smear examinations are still the preferred choices for accurate anemia diagnosis, despite being time-intensive, labor-intensive, and requiring medical expertise. Integrating AI into these invasive diagnostic methods can still be highly useful, as it can reduce the burden on healthcare professionals and help them in diagnostic confirmation.