Abstract

Background: Peripheral blood smear (PBS) examination remains essential in hematological evaluation, providing morphological and diagnostic insight that cannot be fully substituted by automated cell counters. However, manual microscopy is labour-intensive and subject to inter-observer variability. Artificial intelligence (AI)-assisted digital microscopy systems, such as AI100 with Shonit™, aim to enhance reporting efficiency and consistency. Objective: To evaluate the diagnostic performance of AI100 with Shonit™ in comparison with manual microscopy for peripheral blood smear interpretation, focusing on leukocyte differential counts, detection of immature granulocytes and atypical/blast cells, and RBC and platelet morphology assessment. Methods: This cross-sectional study was conducted over ten months at the All India Institute of Medical Sciences, Gorakhpur. A total of 837 peripheral blood smears after exclusion of suboptimal slides were included in the study. Manual microscopy served as the reference standard. The same smears were analysed using AI100 with Shonit™, and results were compared. Pearson correlation, confusion matrix-based diagnostic accuracy, and morphology concordance percentages were calculated. Results: Fair correlations were observed for lymphocytes (r = 0.773, r² = 0.597), eosinophils (r = 0.707, r² = 0.500), and neutrophils (r = 0.698, r² = 0.487), while monocytes showed weaker correlation (r = 0.272, r² = 0.074). Immature granulocyte detection demonstrated sensitivity 81.8% and specificity 75.7%. Atypical/blast detection showed sensitivity 100% and specificity 69.6%. RBC morphology concordance ranged from high (ovalocytes 98.6%, tear drops 95.8%) to moderate (microcytes 56.3%). Platelet adequacy concordance was 66.9%. Conclusion: AI100 with Shonit™ shows strong performance for detection of atypical cells / blasts and Immature Granulocytes, supporting its role as a screening and workload-reduction tool. However, manual confirmation remains essential, particularly for ambiguous morphological abnormalities and platelet assessment. A hybrid AI-assisted workflow provides the most reliable clinical implementation.

Keywords

  • Sigtuple AI100 automation in hematology AI-assisted peripheral blood smear Automated analyzer Image

References

  1. [1] Bain BJ. Diagnosis from the blood smear. N Engl J Med. 2005;353(5):498-507.
  2. [2] Tefferi A, Hanson CA, Inwards DJ. How to interpret and pursue an abnormal complete blood cell count in adults. Mayo Clin Proc. 2005 Jul;80(7):923-36.
  3. [3] Cornet E, et al. CellaVision DM96 evaluation. Int J Lab Hematol. 2008;30(6):536-542.
  4. [4] Adewoyin AS, Nwogoh B. Peripheral blood film review. Ann Ib Postgrad Med. 2014;12(2):71-79.
  5. [5] Kratz A, et al. Automated leukocyte differential imaging. Am J Clin Pathol. 2005;124(5):770-781.
  6. [6] Swolin B, et al. Automated ANN-based differential count. Clin Lab Haematol. 2003;25:139-147.
  7. [7] Amit S, et al. AI-based peripheral smear interpretation. Comput Biol Med. 2024;160:107283.
  8. [8] Cornish RJ, Foran PJ, et al. AI for PBS. Br J Haematol. 2023;200:e52-e64.
  9. [9] Renu Ethirajan and Roja Ramani. “Study on the Performance of SigTuple AI100 in the Analysis of RBC, WBC, Platelet Morphology and WBC Differential Count in a Large Hospital Setup". Acta Scientific Medical Sciences 4.1 (2020): 102-107.
  10. [10] Singh U, Acharya S, Sehgal T. Evaluation of automated blood microscopy system AI100 with ShonitTM in a tertiary care center in Northern India. Int J Innov Res Med Sci. 2024;9(7):422-427.
  11. [11] Chari PS, Prasad S. AI-based PBF analysis pilot. Indian J Hematol Blood Transfusion. 2018;34:125-131.
  12. [12] Aby AE, Salaji S, Anilkumar KK, Rajan T. A review on leukemia detection and classification using artificial intelligence-based techniques. Comput Electr Eng. 2024;118:109-446.
  13. [13] Bengio Y. Deep architectures for AI. Found Trends Mach Learn. 2009;2:1-127.
  14. [14] Dong N, Feng Q, Zhai M, et al. A novel feature fusion based deep learning framework for white blood cell classification. J Ambient Intell Hum Comput. 2023;14:9839-9851. doi:10.1007/s12652-021-03642-7.