Tohoku J. Exp. Med., 2021 July, 254(3)

Deep Learning-Based Nuclear Lobe Count Method for Differential Count of Neutrophils

Mayu Yabuta,1 Iori Nakamura,1 Haruhi Ida,1 Hiromi Masauzi,2 Kazunori Okada,2 Sanae Kaga,2 Keiko Miwa2 and Nobuo Masauzi2

1Graduate School of Health Sciences, Hokkaido University, Sapporo, Hokkaido, Japan
2Faculty of Health Sciences, Hokkaido University, Sapporo, Hokkaido, Japan

Differentiating neutrophils based on the count of nuclear lobulation is useful for diagnosing various hematological disorders, including megaloblastic anemia, myelodysplastic syndrome, and sepsis. It has been reported that one-fifth of sepsis-infected patients worldwide died between 1990 and 2017. Notably, fewer nuclear-lobed and stab-formed neutrophils develop in the peripheral blood during sepsis. This abnormality can serve as an early diagnostic criterion. However, testing this feature is a complex and time-consuming task that is rife with human error. For this reason, we apply deep learning to automatically differentiate neutrophil and nuclear lobulation counts and report the world's first small-scale pilot. Blood films are prepared using venous peripheral blood taken from four healthy volunteers and are stained with MayGrnwald Giemsa stain. Six-hundred 360 ~ 363-pixel images of neutrophils having five different nuclear lobulations are automatically captured by Cellavision DM-96, an automatic digital microscope camera. Images are input to an original architecture with five convolutional layers built on a deep learning neural-network platform by Sony, Neural Network Console. The deep learning system distinguishes the four groups (i.e., band-formed, two-, three-, and four- and five- segmented) of neutrophils with up to 99% accuracy, suggesting that neutrophils can be automatically differentiated based on their count of segmented nuclei using deep learning.

Key words —— blood cell automatic image analysis; computer vision; convolutional neural networks; deep learning; white blood cell morphology


Tohoku J. Exp. Med 2021, 254, 199-206.

Correspondence: Nobuo Masauzi, Faculty of Health Sciences, Hokkaido University, Kita-12-jo, Nishi-5-chome, Kita-ku, Sapporo, Hokkaido 060-0812, Japan.