Investigation of erythrocyte morphologies using artificial intelligence

Introduction: Peripheral blood smear has long been one of the fundamental diagnostic tools in hematology. However, its observer-dependent variability and time-consuming nature limit reliability of this method. In recent years, artificial intelligence (AI)–based image analysis has emerged as an innovative solution to these challenges. This study aimed to evaluate the applicability of AI algorithms in peripheral smear imaging and to investigate the relationship between AI-derived findings and hematological parameters, particularly mean corpuscular volume (MCV). Methods: A total of 1000 peripheral smear images obtained from 286 patients were included in this study. The images were digitized under 100× magnification, and erythrocytes were detected using the YOLOv8 algorithm. In total, 298,043 erythrocytes were analyzed. Results: A moderate, positive, and statistically significant correlation was found between the AI-derived average erythrocyte area and MCV (ρ = 0.4297, p <0.0001). Passing-Bablok regression demonstrated a linear relationship, while Bland-Altman analysis showed a mean difference of –0.37 with limits of agreement ranging from –9.26 to +8.53. CCC values further indicated that the method achieved clinically acceptable concordance. Conclusion and Recommendations: These findings suggest that AI-based analysis of peripheral smears can serve not only as a tool for morphological assessment but also as a complementary method when combined with hematological parameters. This approach may provide valuable contributions, particularly in the early diagnosis of anemia and other hematological disorders. Future large-scale and prospective studies are needed to validate accuracy and strengthen the clinical integration of this method. Keywords: Erythrocyte morphology, Artificial intelligence, Peripheral smear, mean corpuscular volume