Artificial Intelligence in Hematopathology: Evidence that Human-AI Collaboration Outperforms Independent Diagnostic Approaches

Abstract

Artificial Intelligence (AI) has been increasingly integrated into medical diagnostics, particularly in hematopathology. AI has shown promising accuracy in detecting diseases and malignancies in blood and bone marrow. With growing applications in the evaluations of bone marrow, peripheral blood, and lymph node specimens, concerns about AI’s interpretive limitations support its use as a supplement rather than an independent tool. This review examined whether collaboration between AI systems and hematopathologists improves the recognition and diagnosis of bone marrow and lymph node abnormalities compared with either approach alone. We conducted a narrative review, comparing 16 studies. These studies ranged from November 2019 to March 2026. We conducted an organized analysis of focused papers using specific keywords across multiple databases such as Google Scholar and PubMed. These studies were analyzed to evaluate the effects of AI and its integration into hematopathology. From these sources, we highlighted key aspects of AI acting as a tool to enhance hematopathology. The reviewed studies demonstrated that AI systems perform effectively in repetitive pattern-based tasks, differential counting, and disease prediction and classification. Tools such as Cellavision, Morphogo, and other deep learning models improved diagnostic consistency and efficiency. However, AI performance remains limited by reduced generalizability, difficulty detecting rare abnormalities, and challenges in integrating complex clinical information without expert oversight. Across studies, AI demonstrated the greatest benefit when used alongside a hematopathologist rather than as a replacement, performing efficient cell classification, pattern recognition, and analysis of bone marrow samples.