REVIEW OF HAEMATOPATHOLOGY IN THE AGE OF ARTIFICIAL INTELLIGENCE-MACHINE LEARNING BETWEEN 2011 AND 2021

Abstract

Background and Objective: Articial intelligence has transformed pathology diagnostics over the past decade between January 2011 to December 2021, with new emerging technologies and software promising to transform and enhance haematopathology diagnostics further. More rapid and procient AI systems appears to be threatening the role of Haematopathologist in the diagnostic process. This systemic review aims to explore the success of articial intelligence applications in the eld of haematopathology and assess whether the role of haematopathologist will indeed prove redundant in the future. Methods:We performed an extensive search of Pubmed, Medline and National Center for Biotechnology Information (NCBI) at the U.S. National Library of Medicine (NLM) and google scholar databases for articial intelligence in Haematopathology between January 2011 and December 2021.Reference lists of articles were thereafter reviewed for additional reviews. The results are grouped and discussed according to the world health organization grouping of haematopathology disease. Studies where the AI algorithms were compared to that of specialist pathologist were included as this was the main focus and aim of the review. Key content and ndings: Articial intelligent applications on peripheral smears, bone marrow aspirate smears, immunohistochemical stains are documented sequentially in the manuscript from the introduction of whole slide imaging applied to peripheral and bone marrow smears for identication of white blood cells to the application of more complex convoluted neural networks for discrimination of lymphoma and leukaemia subtypes and lymphoma grading. All the studies documented in this review have shown favourable outcome for articial intelligence applications to haematopathology disease. Conclusion: The above studies have demonstrated that articial intelligence can be successfully integrated into haematopathology diagnostics. Although all studies were shown to be comparable to the pathologist, there is a requirement for further standardisation and validation studies for optimization of deep learning algorithms. The notion that AI will replace the pathologist is also incorrect. The microscope will not be replaced. Rather, AI integration into pathology is meant enhance the accuracy and speed of diagnostic workows enabling the pathologist to focus on more complex laboratory problems. AI and human pathologists should co- operate, rather than compete.