Artificial Intelligence based on identification of red blood cells morphological abnormalities in anaemia types : a scoping review

Abstract

Anaemia is the most common type of disorder in the world today. Conventionally, manual microscopic system performs peripheral blood smear to detect morphologically abnormal red blood cells in different anaemia types. Currently, Artificial Intelligence based on machine learning and neural networks are becoming a standard method for recognizing morphological features of red blood cells in medical field. It uses deep learning to recognize images. AI models have been trained and validated for image processing in clinical haematology laboratories. The objective of this study is to describe and differentiate AI systems and manual methods for diagnosing RBC morphology in different types of anaemia. This study was a scoping review. Totally 20 articles have been selected for this study which included PubMed and ScienceDirect online databases. Analysis table was prepared for data analysis and was used to build up results by an Excel spreadsheet under the content analysis method. This study has shown high accuracy and sensitivity for AI based systems rather than manual method for detecting anaemic red blood cells such as sickle cells, microcytes, elliptocytes, target cells, pencil cells, helmet cells in different anemia types like sickle cell anaemia, haemolytic anaemia and iron deficiency anaemia. According to this study, AI system has high accuracy, sensitivity and efficacy when com paring with manual methods for identifying morphologically abnormal red blood cells in anaemic patients. Also out of AI models, deep neural network has shown more accuracy and sensitivity.