Identification of fish blood cell populations on the basis of a convolutionary neural network for compiling a leukogram

In aquaculture, physiological assessment is required to monitor the health status of fish. Blood is the most responsive system in the organism of hydrobionts to changes in external factors. The study of hematological parameters of fish allows for early diagnosis of diseases, working out the technological mode of breeding and rearing, and selection. The typing of cells in circulating fluids is important for compiling hemocytic and leukocyte formulas characterizing the cellular component of the organism’s immune response.

In the present study, convolutional neural network models are developed to classify blood cells of carp and sturgeon fish. The quality of the models is estimated based on the metrics Accuracy and Precision, Recall, F₁ with macro-averaging.

Based on the processing of blood images, 1104 images of blood cells of carp and sturgeon fish were prepared, including 15 cell populations: hemohistoblasts, myeloblasts, promyelocytes, myelocytes, metamyelocytes, rod-shaped neutrophils, segmented neutrophils, eosinophils, basophils, monocytes, lymphocytes, erythroblasts, normoblasts, mature erythrocytes, and platelets.

Models of a convolutional neural network have been developed to recognize populations of blood cell elements (erythrocytes, leukocytes, platelets) of carp and sturgeon fish. The models were trained on 80% of the prepared images, avoiding the problem of overtraining, as evidenced by the constructed graphs of the loss function (sparse categorical cross entropy) and accuracy during the learning process.

The constructed models make it possible to recognize blood cells of carp fish with an accuracy of 75.0% (metric F₁ with macro-averaging is 0.570) and blood cells of sturgeon fish with an accuracy of 76.6% (F₁ with macro-averaging is 0.664).

1. Dolgov V.V., Menshikov V.V. Clinical laboratory diagnostics. National leadership // Moscow: GEOTAR-Media. – 2016. - 688s. (In Russian)

2. Zhiteneva, L.D., Makarov E.V., Rudnitskaya O.A. Fundamentals of ichthyohematology (in comparative aspect). – Rostov-on-Don: Everest, 2004. – 312 p. (In Russian)

3. Ivanov A.A., Pronina G.A., Koryagina N.Yu. Hematology of poikilothermic hydrobionts: Monograph. – Irkutsk: Megaprint LLC, 2018. – 133 p. (In Russian)

4. Maslova N.I., Petrushin A.B. Methodological recommendations for the management of the breeding process in fish farming (on the example of carp). – M.: Russian Agricultural Academy, 2005. – 27c. (In Russian)

5. Patterson J., Gibson A. Deep learning from the point of view of practice. – Moscow: DMK Press, 2018. – 418 p. (In Russian)

6. Pishchenko E.V. Hematology of freshwater fish: Textbook. – Novosibirsk: Novosibirsk State Agrarian University. un-t, 2002. – 48s. (In Russian)

7. Pronina G.I., Ivanov A.A., Mannapov A.G., Sanaya O.V. Immunity of poikilothermic hydrobionts // Izvestiya TSKHA. – 2021. – No. 2. – pp. 71-91. (In Russian)

8. Pronina G.I. Clinical laboratory diagnostics. Workshop: Textbook for universities. – M., 2021. – 88 p. (In Russian)

9. Pronina G.I., Koryagina N.Yu. Methodology of physiological and immunological assessment of hydrobionts: Textbook. – St. Petersburg: Lan, 2017. – 96c. (In Russian)

10. Agustin, R.I., Arif A., Sukorini U. Classification of immature white blood cells in acute lymphoblastic leukemia L1 using neural networks particle swarm optimization // Neural Comput & Applic 33, 10869–10880, 2021. – URL: https://doi.org/10.1007/s00521-021-06245-7

11. Fernández J.C., Carbonero M., Gutiérrez P.A. et al. Multi-objective evolutionary optimization using the relationship between F1 and accuracy metrics in classification tasks // Appl Intell 49, 2019, 3447–3463. – URL: https://doi.org/10.1007/s10489-019-01447-y

12. Hameed Z., Garcia-Zapirain B., Aguirre J.J. et al. Multiclass classification of breast cancer histopathology images using multilevel features of deep convolutional neural network // Sci Rep 12, 15600, 2022. – URL: https://doi.org/10.1038/s41598-022-19278-2

13. Le Floc’h N., Gondret F., Resmond R. Identification of blood immune and metabolic indicators explaining the variability of growth of pigs under contrasted sanitary conditions. BMC Vet Res 17, 166 (2021). https://doi.org/10.1186/s12917-021-02872-3

14. Mortaz E. Imbalance accuracy metric for model selection in multi-class imbalance classification problems // Knowledge-Based Systems, Volume 210, 2020, 106490. – URL: https://doi.org/10.1016/j.knosys.2020.106490

15. Sadafi A., Shen D. Multiclass Deep Active Learning for Detecting Red Blood Cell Subtypes in Brightfield Microscopy // Medical Image Computing and Computer Assisted Intervention – MICCAI 2019. Lecture Notes in Computer Science(), vol. 11764. Springer, Cham, pp. 685-693, 2019. – URL: https://doi.org/10.1007/978-3-030-32239-7_76

16. Shilaskar S., Ghatol A. Diagnosis system for imbalanced multi-minority medical dataset // Soft Comput 23, 4789–4799, 2019. – URL: https://doi.org/10.1007/s00500-018-3133-x

17. Sokolova M., Lapalme G. A systematic analysis of performance measures for classification tasks // Information Processing & Management, Volume 45, Issue 4, Pages 427-437, 2009. – URL: https://doi.org/10.1016/j.ipm.2009.03.002

18. Takahashi K., Yamamoto K., Kuchiba A. Confidence interval for micro-averaged F1 and macro-averaged F1 scores // Appl Intell 52, 4961–4972, 2022. – URL: https://doi.org/10.1007/s10489-021-02635-5

19. tf.keras.callbacks.EarlyStopping. – Text: electronic // tensorflow.org – the official website of TensorFlow open source software library for machine learning. – URL: https://www.tensorflow.org/api_docs/python/tf/keras/callbacks/EarlyStopping

20. Upadhyay V., Rathod D. Location-Based Crime Prediction Using Multiclass Classification Data Mining Techniques // (eds) IOT with Smart Systems. Smart Innovation, Systems and Technologies, vol 251. Springer, Singapore, pp. 619-626, 2022. – URL: https://doi.org/10.1007/978-981-16-3945-6_61