Is Artificial Intelligence Ready to Replace the Clinician in Formula Leukocyte Examination with Peripheral Spread?

BACKGROUND: The analysis of peripheral blood smear (PBS) is a routine test that plays an important role in disease diagnosis. Automation of blood analysis will not only save time and money while reducing errors, but it will also be useful and instructive for frontline workers, especially during pandemics when access to hematologists is limited. METHODS: Mantiscope is a cloud-based slide scanner device with useful possibilities that can collect images from PBS with autofocus. It is integrated into a platform that uses artificial intelligence to collect and analyze the desired number of images on the samples. RESULTS: One hundred and three PBS was evaluated by three pediatric hematologists using light microscopy. A comparison was made between the created digital dataset and the artificial intelligence algorithm trained in peripheral spread assessment using the intraclass correlation coefficient (ICC). When the correlation coefficients for neutrophils, lymphocytes, monocytes, eosinophils, and basophils were compared using the means of all three evaluators and the artificial intelligence algorithm, the results were 0.885, 0.815, 0.042, and 0.599, respectively. As no blasts were detected in peripheral smears by the artificial intelligence algorithm, no comparison with evaluators was possible. CONCLUSIONS: Mantiscope system will help primary care physicians manage viral and bacterial infections because of the excellent reliability found in neutrophil and lymphocytic series evaluations using this developed algorithm. However, since it was unable to detect cells from the blast line, it was thought that it would not be adequate for the differential diagnosis of malignancy and that suspicious patients should be referred to a hematologist.