Morphological Analysis on Peripheral Blood
A paper presented at the Korean Society for Laboratory Hematology by Seoul St. Mary’s Hospital
Published in 2023.01
Summary
[Background]
The UIMD PBIA is a newly developed automated image analysis device for peripheral blood cells. In this study, the accuracy and processing speed of the UIMD PBIA for white blood cell (WBC) classification were evaluated.
[Methods]
A total of 29,605 white blood cell images obtained from 242 peripheral blood smears, including 192 samples from patients with abnormal findings and 50 samples from healthy individuals without abnormal findings, were analyzed to evaluate the accuracy and processing speed of the device.
[Results]
The UIMD PBIA demonstrated an accuracy of 99% in samples from healthy individuals and an overall accuracy of 99.2% in five-part white blood cell differential counting.
Misclassification occurred most frequently in immature granulocytes, blasts, and abnormal lymphocytes, for which the classification accuracy ranged from 81% to 93.9%.
Abnormal blood cells tended to be misclassified as other cells within the same lineage.
The processing speed of the device was 42 slides per hour, and 29 slides per hour in cases with cytopenia.
[Conclusion]
The UIMD PBIA provides rapid and accurate white blood cell classification results and is expected to be particularly useful in cases with normal findings or cytopenia.
Morphological Analysis on Peripheral Blood
A paper presented at the Korean Society for Diagnostic Hematology by Inje University Busan Paik Hospital
Published in 2024.10
Summary
[Background]
The use of digital cell morphology analyzers has increased in recent years to overcome the limitations of manual white blood cell (WBC) counting. In this study, we evaluated the performance of the newly developed digital morphology analyzer, UIMD PBIA, using samples with leukopenia.
[Methods]
A total of 159 leukopenia samples were classified into 4 groups according to the degree of leukopenia.
(Group 1: WBC≤1.0×109/L, Group 2: 1.0–2.0×109/L, Group 3: 2.0–3.0×109/L, Group 4: 3.0–4.0×109/L).
Using 23,358 WBC images obtained from these samples, the accuracy and processing speed of the UIMD PBIA were analyzed and compared.
[Results]
The UIMD PBIA demonstrated an overall accuracy of 97.5%, showing more than 90% accuracy across all cell groups except abnormal lymphocytes.
Blasts showed a statistically significant correlation between the manual method and UIMD PBIA (r > 0.7).
When both examiners reported blast or immature granulocyte counts exceeding 1%, the UIMD PBIA showed a 100% positive agreement with the manual method. The processing time of the UIMD PBIA was shorter than that of the manual method in Groups 1 and 2 compared to Groups 3 and 4.
[Conclusion]
Although misclassification of abnormal cells still requires further improvement, the UIMD PBIA provides rapid and accurate WBC differential results in leukopenic samples. When selectively applied according to WBC count and clinical diagnosis, the UIMD PBIA may serve as a useful tool in clinical practice.
Morphological Analysis on Peripheral Blood
A paper presented at the International Conference by Soonchunhyang University Cheonan Hospital
Published in 2025.01
Summary
[Background]
The performance of the UIMD PBIA, an automated digital morphology analyzer using deep learning for white blood cell (WBC) classification in peripheral blood smears, was evaluated and compared with that of the commercially available DI-60 (Sysmex).
[Methods]
A total of 461 PBS slides were pre-classified and analyzed using both the PBIA and DI-60. The pre-classification performance of each device was evaluated based on post-classification results verified by users. The mean differences among cell classes were calculated for both devices, and the correlations between pre- and post-classifications as well as manual counts were assessed for each cell type.
[Results]
Overall, the pre-classification performance of the PBIA was superior to that of the DI-60 across most cell classes, demonstrating higher accuracy, lower FPR and FNR, and stronger agreement. The PBIA showed an accuracy of over 90.0% in all cell classifications and a Cohen’s kappa value of 0.934, which was higher than that of the DI-60 (accuracy 45.5%, kappa 0.629).Although the pre-classification performance of both devices declined when abnormal cells were observed in manual counts, the PBIA still demonstrated superior performance.
[Conclusion]
Under the conditions of this study, the PBIA demonstrated superior pre-classification performance compared to the DI-60, showed a stronger correlation with manual counts, and exhibited less discrepancy between pre- and post-classification.
Although morphological examination of blood cells inherently involves subjective interpretation and requires highly experienced technicians, the high accuracy of the PBIA may help reduce diagnostic errors of clinically significant abnormal cells when used by less experienced users. In conclusion, the PBIA outperformed the DI-60, highlighting its strong clinical utility.
Morphological Analysis on Bone Marrow
A paper presented at the LMCE Conference by Korea University Hospital
Published in 2026.02
Summary
[Background]
Bone marrow aspirate (BMA) differential counts are essential for diagnosing hematologic disorders. Although manual microscopic examination remains the standard method, it is labor-intensive and susceptible to inter-observer variability. Several AI-assisted systems have been developed to improve efficiency and consistency; however, few have been validated for routine clinical use. We evaluated BMIA-12A, a deep learning–based system for automatic classification of bone marrow cells.
[Results]
BMIA-12A achieved an accuracy of more than 93% for both wedge and squash smear preparations. All cell categories demonstrated sensitivity and specificity above 90%. A strong correlation between AI pre-classification and expert review was observed (wedge r = 0.996, squash r = 0.993; P < 0.001). Overall, BMIA-12A showed clinically acceptable concordance with expert assessments, indicating its potential usefulness as an assistive tool in BM evaluation.
[Conclusion]
BMIA-12A demonstrated strong agreement with expert review, achieving over 93% accuracy and over 90% sensitivity and specificity across major cell types. Although limited representation of rare categories remains a challenge, overall performance was clinically acceptable. These findings suggest that BMIA-12A can serve as a reliable adjunct in bone marrow smear evaluation and that further validation across diverse clinical settings will help establish its broader applicability.
Login with Naver ID 
Login with Kakao ID 
Login with Google ID