Comprehensive review on learning models of leukemia detection based on morphological information

Leukemia is a blood cancer caused by the abnormal multiplication of immature white blood cells and is the 11th most prevalent cancer worldwide. Early and accurate leukemia diagnosis is critically important because it directly influences treatment decisions and patient outcomes. Delayed or inaccurate diagnosis leads to disease progression, a higher risk of life-threatening infections, and poorer survival rates. Conventional detection techniques, including manual microscopic analysis and complete blood count tests, are time-consuming, less reliable, and require medical experts. Accurate and timely detection improves treatment outcomes and increases survival chances. Various artificial intelligence approaches have been implemented for leukemia detection and offer satisfactory results. This study presents a systematic review of artificial intelligence-based approaches used in leukemia diagnosis. It includes an analysis of image acquisition methods such as peripheral blood smear microscopy, flow cytometry, bone marrow biopsy imaging, and advanced imaging techniques. Preprocessing steps such as noise removal, artifact removal, and image enhancement methods that improve image contrast and quality are discussed. The review provides an overview of segmentation techniques, from conventional clustering algorithms to deep learning frameworks. It also examines machine learning and deep learning models used in leukemia classification, highlighting their merits and challenges while identifying existing issues and potential future directions.