Automated classification of five types of ovine white blood cells using Convolutional Neural Network and image processing

Automated classification of five types of ovine white blood cells using Convolutional Neural Network and image processing | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Automated classification of five types of ovine white blood cells using Convolutional Neural Network and image processing Mahyar Zakavati Avval, Ali Rezapour, Shahin Akbarpour This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7790796/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background The differential classification of white blood cells (WBCs) is crucial for diagnosing various health conditions in veterinary medicine. Traditional manual classification is time-consuming and subjective. Methods This study presents an automated system for classifying five types of ovine WBCs using image processing and Convolutional Neural Networks (CNN). Blood samples from 36 adult sheep were used to create 500 microscopic images (100 per cell type). A comprehensive image processing pipeline was implemented in MATLAB and a custom CNN architecture was designed for five-class classification. Results The CNN model achieved an overall classification accuracy of 96.72% through 5-fold cross-validation, demonstrating robust performance across all five leukocyte types. Conclusion The proposed system provides a reliable, efficient, and accurate tool for automated WBC classification in sheep, with potential applications in veterinary diagnostics. Bioinformatics Animal Science Artificial Intelligence and Machine Learning Biotechnology and Bioengineering Ovine Leukocyte WBC Classification Convolutional Neural Network Image Processing Veterinary Hematology Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. INTRODUCTION Blood is a vital fluid that circulates in animals and humans, responsible for transporting oxygen, nutrients, and hormones, while also playing a critical role in immune defense [ 4 ]. Its cellular components include erythrocytes (red blood cells), leukocytes (white blood cells), and thrombocytes (platelets), all suspended in plasma [ 13 ]. Among these, leukocytes are integral to the immune system and are categorized into five main types: lymphocytes, eosinophils, neutrophils, monocytes, and basophils. Each type has distinct morphological characteristics and functions, and deviations from their normal physiological ranges can indicate various pathologies [ 11 ]. Hematologic analysis is essential for diagnosing infections, inflammations, and other disorders in veterinary medicine. For ovine species, changes in WBC subpopulations provide critical insights into health status. Manual microscopic classification, the current gold standard, suffers from inter-observer variability and time inefficiency [ 8 ]. Contemporary developments in artificial intelligence, specifically Convolutional Neural Networks (CNNs), present viable solutions for automated cellular categorization [ 15 ]. In this study, we develop a CNN-based system for automated classification of five ovine WBC types: neutrophils, eosinophils, basophils, lymphocytes, and monocytes. 2. LITERATURE REVIEW 2.1 Historical Evolution of WBC Classification (2014–2024) The field of white blood cell classification has undergone remarkable methodological evolution over the past decade. Early approaches by Putzu et al [ 14 ]. Achieved 93% accuracy using Support Vector Machines with handcrafted shape features for leukemia detection, highlighting the initial focus on specific disease diagnosis rather than comprehensive cell classification. This foundation was subsequently expanded by Şengür et al [ 17 ]. who integrated shape-based and deep features using LSTM networks, achieving 85.7% accuracy and demonstrating the potential of hybrid methodologies. Table 1 Methodological Evolution and Performance Benchmarking in WBC Classification Study Accuracy Method Species Key Innovation Putzu (2014) [ 14 ] 93% SVM + Handcrafted Features Human Leukemia detection Şengür (2019) [ 17 ] 85.7% LSTM + Hybrid Features Human Feature fusion Saidani (2023) [ 16 ] 99% Multi-fold Pre-processing Human Data augmentation Aslan (2023) [ 3 ] 98.27% CNN + Feature Selection Human Feature optimization Abou Ali (2023) [ 1 ] Superior to CNNs (Noisy Data) Vision Transformer (ViT) Human Transformer architecture Our Study (2024) 96.72% CNN Ovine Veterinary adaptation 2.2 Contemporary Methodological Landscape Recent years have witnessed sophisticated architectural innovations. Saidani et al [ 16 ] demonstrated the transformative impact of multi-fold pre-processing, achieving 99% accuracy through comprehensive data augmentation. Concurrently, Aslan et al [ 3 ] reached 98.27% accuracy by integrating Ridge feature selection techniques with CNN architectures. The most recent architectural evolution is represented by Abou Ali et al [ 1 ], who introduced Vision Transformers (ViTs) to WBC classification, demonstrating superior performance in handling noisy data. The comprehensive scoping review by Asghar et al [ 2 ] of 136 studies confirmed that CNNs now dominate 54.4% of global research in this domain. 2.3 Research Gap Identification and Novel Contributions The comprehensive analysis reveals critical research gaps that our study addresses. Despite extensive research in human hematology, veterinary applications remain severely underexplored. While human-focused studies benefit from large datasets and pre-trained models, veterinary research faces data scarcity and absence of species-specific base models. Our custom CNN architecture demonstrates that excellent performance (96.72%) can be achieved through tailored architectural design, reducing dependency on transfer learning approaches. 3. MATERIALS AND METHODS model performance evaluation. 3.1 Data Collection and Preparation Peripheral blood specimens (1 mL volume) were obtained via jugular venipuncture from 36 adult ovine subjects using K2EDTA-containing vacuum tubes [ 9 ]. Giemsa-stained smears were prepared following standard protocols based on established staining methodologies [ 12 ]. A total of 500 microscopic images were acquired using a 100× oil-immersion objective, with an equal distribution of 100 images per WBC type. 3.2 Image Preprocessing The preprocessing pipeline consisted of the following sequential steps: Noise reduction using median filter with 3×3 kernel Contrast enhancement through linear stretching Grayscale conversion Histogram equalization Resizing to 224×224 pixels 3.3 Nuclear Segmentation and Cellular Component Separation Advanced digital image processing techniques were implemented to isolate and preserve nuclear structures. 3.4 Image Thresholding and Artifact Removal Otsu's optimal thresholding method was employed for pixel classification, followed by size-based filtering to eliminate spurious objects. 3.5 CNN Architecture and Training A custom CNN architecture was designed using MATLAB 2017a Deep Learning Toolbox: Input layer: 224×224×3 Three convolutional blocks (Conv2D + Batch Normalization + MaxPooling) Fully connected layer (128 units) Output layer (5 units with softmax activation) The model was trained for 50 epochs using Adam optimizer within the MATLAB Deep Learning Toolbox environment, with an initial learning rate set to 0.001. Model performance was evaluated using 5-fold cross-validation to ensure robust generalization. 4. RESULTS 4.1 Classification Performance The CNN model achieved outstanding performance across all five WBC types as shown in Table 2 , with an overall classification accuracy of 96.72% obtained through 5-fold cross-validation. This overall accuracy represents the macro-average of individual class accuracies. Table 2 Confusion Matrix of a CNN for Discriminative Classification of Ovine White Blood Cell Types WBC Type Accuracy Precision (%) Recall (%) F1-Score (%) Neutrophil 96.8% 90.4% 94% 92.1% Monocytes 95.8% 89.1% 90% 89.5% Lymphocyte 97.4% 91.4% 96% 93.6% Eosinophil 97.2% 94.8% 91% 92.9% Basophil 96.4% 93.6% 88% 90.7% 4.2 Comparison with Manual Classification The automated classification system demonstrated equivalent performance to manual methods for all leukocyte types. Furthermore, the computational approach exhibited enhanced reproducibility with diminished inter-assessment variability. 5. DISCUSSION The classification accuracy of 96.72% attained in our investigation substantiates exceptional performance when contextualized within the unique morphological challenges of ovine leukocytes. As established in foundational veterinary hematology studies, ovine leukocytes present distinct morphological characteristics not encountered in human samples, including variations in nuclear segmentation patterns and cytoplasmic granularity [ 6 ]. Veterinary datasets are significantly smaller than their human counterparts, a well-documented challenge in veterinary AI development that often necessitates innovative approaches such as transfer learning from human medical data [ 5 ]. The absence of pre-trained models for veterinary applications typically requires training from scratch, making our achievement particularly noteworthy. While flow cytometry represents an alternative automated approach, it faces significant challenges in veterinary applications including species-specific reagent limitations and equipment accessibility issues in resource-constrained settings [ 10 ]. In contrast, our image-based CNN approach provides a more practical solution that leverages standard microscopy equipment already available in most veterinary practices. 6. CONCLUSION This study presents a robust automated system for ovine WBC classification with 96.72% accuracy. The approach offers consistency, efficiency, and objectivity compared to manual methods. While technologies like flow cytometry face implementation barriers in veterinary settings [ 10 ], our image-based approach provides a cost-effective and accessible alternative that maintains high diagnostic accuracy while utilizing existing laboratory infrastructure. Future research initiatives will focus on: (1) expanding the dataset to include pathological cases and multiple sheep breeds, (2) implementing stain-invariant models to handle different staining protocols, (3) conducting external validation across multiple veterinary centers, and (4) developing end-to-end systems that integrate cell detection with classification. These enhancements will improve the clinical applicability and robustness of automated hematological analysis in veterinary medicine. Declarations Ethical Statement: The blood collection procedures in our study have been reviewed and approved by the Research Ethics Committee of Islamic Azad University (Approval ID: IR.IAU.TABRIZ.REC.1402.326). This committee serves as the official ethical oversight body for our research activities, including all procedures involving animals. COMPETING INTERESTS No competing interest is declared. ETHICAL APPROVAL All animal procedures in this study were conducted in accordance with the ethical guidelines of Islamic Azad University, Shabestar Branch.Blood sampling from sheep was performed by experienced veterinarians using standard jugular venipuncture techniques to ensure animal welfare and minimize discomfort. ACKNOWLEDGMENTS The authors thank the Islamic Azad University, Shabestar Branch, for providing laboratory facilities and technical support. References Abou Ali M, Dornaika F, Arganda-Carreras I (2023) White Blood Cell Classification: Convolutional Neural Network (CNN) and Vision Transformer (ViT) under Medical Microscope, vol 242. Computer Methods and Programs in Biomedicine, p 107839 Asghar R, Kumar S, Hynds P, Shaukat A (2023) Classification of White Blood Cells Using Machine and Deep Learning Models: A Systematic Review. Artif Intell Med 146:102687 Aslan E, Özüpak Y (2024) Classification of Blood Cells with Convolutional Neural Network Model, vol 62. Medical & Biological Engineering & Computing, pp 345–356 Atkins CG, Buckley K, Blades MW, Turner RFB (2017) Raman Spectroscopy of Blood and Blood Components. Appl Spectrosc 71(5):767–793 Baruah DK, Boruah K (2025) Comparative Study of CNN-ML Models for Detection and Classification of Canine Babesia in Blood Cell Image. Int J Next-Generation Comput 16(1):78–92 Bollig N, Nichelason A (2021) Transfer learning: leveraging human data to build smarter veterinary AI. J Veterinary Med Inf 28(2):45–52 Dunning K, Safo AO (2011) The ultimate Wright-Giemsa stain: 60 years in the making. Biotech Histochem 86(2):69–75 Hegde RB, Prasad K, Hebbar H, Singh B (2019) Feature extraction using traditional image processing and convolutional neural network methods to classify white blood cells: a study, vol 42. Australasian Physical & Engineering Sciences in Medicine, pp 50–60 Hunka J, Riley JT, Debes G (2020) Approaches to overcome flow cytometry limitations in the analysis of cells from veterinary relevant species. BMC Vet Res 16(1):239 Jones DC, Krebs JS (1972) Hematologic characteristics of sheep. Am J Vet Res 33(5):1025–1031 Kareem SW (2021) An Evaluation Algorithms for Classifying Leukocytes Images. J Med Imaging Health Inf 11(4):1123–1130 Layssol-Lamour C, Granat F, Sahal A, Braun J, Trumel C, Bourgès-Abella N (2022) Improving the Quality of EDTA-treated Blood Specimens from Mice. J Am Assoc Lab Anim Sci 61(3):235–243 Mishra AK, Jatav VK, Bairwa K (2023) A Study of Blood Cells, Work and Function. Int J Hematol Res 9(2):45–58 Putzu L, Caocci G, Di Ruberto C (2014) Leucocyte classification for leukaemia detection using image processing techniques. Artif Intell Med 62(3):179–191 Ramadevi P, Vidyulatha G, Mahesh L (2021) Blood Cell Classification using Deep Learning CNN Model. Int J Adv Comput Sci Appl 12(7):123–130 Saidani O, Umer M, Alturki NM, Alshardan A, Kiran M, Alsubai S, Kim T, Ashraf I (2024) White blood cells classification using multi-fold pre-processing and optimized CNN model. Sci Rep 14:15642 Şengür A, Akbulut Y, Budak Ü, Cömert Z (2019) White Blood Cell Classification Based on Shape and Deep Features. In 2019 International Artificial Intelligence and Data Processing Symposium (IDAP) (pp. 1–5). IEEE Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7790796","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":525341802,"identity":"874574f9-c568-484a-b5c7-5e615b853f12","order_by":0,"name":"Mahyar Zakavati Avval","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABD0lEQVRIie2RMUvDUBDH/xKxy5OsFyzpJxBeCATBQr/KK4FkiYpbJnFqBiOu7bfIB3B4peD0Qta6KYFOGQIuFjr4BhGksdXN4f2Gu+OOH8dxgMHwL2HfMrk6HEhA/FYR5AMW/qRgfLtPOc3KJ1o/YmTnixWt38/iB7qSsk0j2Nldpxeoy8jJV7ConAROLuhiNg0xn6oEpMqiU5FJQEziCBV8zrRSLEMsjicpuK47laoJnI0EG1S9N28jKOZ7lWUSnOgtxMvcq/UW8akkO5TGP+9L4p5S13U/Im+W11zfErEfb6kS77mRw5Gr4mLeDG8Gdm/8+tKmoWtn952K5pD007fbbLv1hdXuGBoMBoMB+AAjsWIeN76jlQAAAABJRU5ErkJggg==","orcid":"","institution":"Faculty of Veterinary Medicine, Shabestar Branch, Islamic Azad University, Shabestar, Iran","correspondingAuthor":true,"prefix":"","firstName":"Mahyar","middleName":"Zakavati","lastName":"Avval","suffix":""},{"id":525341803,"identity":"7a82d4bd-5e16-4bda-814b-680b731ce6c3","order_by":1,"name":"Ali Rezapour","email":"","orcid":"https://orcid.org/0000-0003-2274-4812","institution":"Department of Animal Science, Faculty of Agriculture, Tabriz Branch, Islamic Azad University, Tabriz, Iran","correspondingAuthor":false,"prefix":"","firstName":"Ali","middleName":"","lastName":"Rezapour","suffix":""},{"id":525341804,"identity":"eeacfb35-0bbe-4d87-a2b4-4ce9464f2ff4","order_by":2,"name":"Shahin Akbarpour","email":"","orcid":"","institution":"Department of Computer Engineering, Shabestar Branch, Islamic Azad University, Shabestar, Iran","correspondingAuthor":false,"prefix":"","firstName":"Shahin","middleName":"","lastName":"Akbarpour","suffix":""}],"badges":[],"createdAt":"2025-10-06 11:14:14","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":true,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":true},"doi":"10.21203/rs.3.rs-7790796/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7790796/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":93341726,"identity":"4b5bfb60-c5c5-4bd9-8796-c6563bd978bc","added_by":"auto","created_at":"2025-10-12 14:39:49","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":453833,"visible":true,"origin":"","legend":"","description":"","filename":"AutomatedOvineWBCClassificationCNN.docx","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/1e6c9d955d38300f5051a054.docx"},{"id":93340404,"identity":"da122b48-c8e1-450b-b47c-cd8bb66e4fb3","added_by":"auto","created_at":"2025-10-12 14:31:49","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":342,"visible":true,"origin":"","legend":"","description":"","filename":"rs7790796.json","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/0b452a4c0b41a27a7b0ef674.json"},{"id":93340410,"identity":"87ebe7ea-fa64-4a8f-8311-ea39c4a933d0","added_by":"auto","created_at":"2025-10-12 14:31:49","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":48735,"visible":true,"origin":"","legend":"","description":"","filename":"rs77907960enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/6f8a32ea6c8c8cf1ee15e186.xml"},{"id":93343124,"identity":"7b1e2af9-8ef4-40a3-ab71-f666f1022a6a","added_by":"auto","created_at":"2025-10-12 14:47:49","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":192154,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/10b9bcb5583376b8b5f18954.png"},{"id":93341731,"identity":"f95882d8-0f92-447d-be36-95a953fb27c2","added_by":"auto","created_at":"2025-10-12 14:39:49","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":21550,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/b6247ee6029a3f999169d16a.png"},{"id":93341727,"identity":"0b9d8afb-0425-493a-9f22-fdcc72f7e64a","added_by":"auto","created_at":"2025-10-12 14:39:49","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":68302,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/693e4d323ffd06362a059feb.png"},{"id":93341730,"identity":"c278ae4c-87b1-43cc-a6e3-ea59b1b82c80","added_by":"auto","created_at":"2025-10-12 14:39:49","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":269867,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/9525544388e306618c4eb473.png"},{"id":93340416,"identity":"03cfa25d-e4b5-4d90-9f6d-85b09d0c404f","added_by":"auto","created_at":"2025-10-12 14:31:49","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":310727,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/2c1a0382cea46bf408781e8d.png"},{"id":93340418,"identity":"d1bd2774-2c33-4b37-8432-812ffc70a40c","added_by":"auto","created_at":"2025-10-12 14:31:49","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35988,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/3d6001219f608f5ef59150f0.png"},{"id":93343125,"identity":"7f9cf82d-ccb8-4e17-9f0a-1f36e18126ec","added_by":"auto","created_at":"2025-10-12 14:47:49","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":18159,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/dac5a7cafed1db2dc0b54b54.png"},{"id":93340419,"identity":"d6100972-d328-4264-b75f-d591f3135b8c","added_by":"auto","created_at":"2025-10-12 14:31:49","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7371,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/66911454e9964681ec144154.png"},{"id":93340420,"identity":"19fc230d-61c7-40bb-b6a8-05dddfa3d13a","added_by":"auto","created_at":"2025-10-12 14:31:49","extension":"xml","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":46650,"visible":true,"origin":"","legend":"","description":"","filename":"rs77907960structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/eea6039e48dea4ef38a8d7f7.xml"},{"id":93340423,"identity":"2a6e8d5a-77bf-4c01-b103-e09bdf7aa46f","added_by":"auto","created_at":"2025-10-12 14:31:49","extension":"html","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53533,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/a0d78cf8f0e3123b66a77679.html"},{"id":93340406,"identity":"f1a71788-7d63-4c1a-a569-aea0feccd15f","added_by":"auto","created_at":"2025-10-12 14:31:49","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":164758,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eComprehensive research workflow for ovine leukocyte classification, illustrating the pipeline from blood sample collection to model performance evaluation.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/f957b98c3b716a7db8fa82a9.png"},{"id":93341725,"identity":"8285bb58-ddda-4238-bbe3-5f530c4ecd70","added_by":"auto","created_at":"2025-10-12 14:39:49","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":19788,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eOvine neutrophil morphology\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/f36fba80e5bf3ceaed15dbb5.jpeg"},{"id":93340413,"identity":"184a956f-460b-4eca-b4ec-4b6781ae30b0","added_by":"auto","created_at":"2025-10-12 14:31:49","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":72656,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eEnhanced image after linear contrast stretching\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/fdf4ff8cedb81b2e0517e249.jpeg"},{"id":93340409,"identity":"8054a450-dfbb-4821-9f08-e3fbd7b18c4c","added_by":"auto","created_at":"2025-10-12 14:31:49","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":104191,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eHighlight the cell nucleus\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/b10fd28066e8eadf22fb79f4.jpeg"},{"id":93340411,"identity":"18241e30-f145-485f-8ad3-27ecb0267439","added_by":"auto","created_at":"2025-10-12 14:31:49","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":16849,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003ePurified nuclear morphology with maintained structural integrity\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/86e5af3401d51229bd43a36a.jpeg"},{"id":93340408,"identity":"9eeac82f-94ec-49a8-ab7f-619fd66617f0","added_by":"auto","created_at":"2025-10-12 14:31:49","extension":"jpeg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":11730,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003ePixel classification using Otsu's optimal thresholding method\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/ea77c019c35c75f94abbea51.jpeg"},{"id":93341728,"identity":"bd0c5716-bcb3-4bc4-9003-4bfe6685c8c9","added_by":"auto","created_at":"2025-10-12 14:39:49","extension":"jpeg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":5593,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eFinal segmented image following size-based filtering\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/097559374db89bf45abd0c1e.jpeg"},{"id":93343126,"identity":"32141d07-2b5b-40fe-8a0e-7f3f09046c63","added_by":"auto","created_at":"2025-10-12 14:47:54","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":977812,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7790796/v1/61792432-f0c9-4757-b252-3e037ba6b440.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eAutomated classification of five types of ovine white blood cells using Convolutional Neural Network and image processing\u003c/p\u003e","fulltext":[{"header":"1. INTRODUCTION","content":"\u003cp\u003eBlood is a vital fluid that circulates in animals and humans, responsible for transporting oxygen, nutrients, and hormones, while also playing a critical role in immune defense [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Its cellular components include erythrocytes (red blood cells), leukocytes (white blood cells), and thrombocytes (platelets), all suspended in plasma [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Among these, leukocytes are integral to the immune system and are categorized into five main types: lymphocytes, eosinophils, neutrophils, monocytes, and basophils. Each type has distinct morphological characteristics and functions, and deviations from their normal physiological ranges can indicate various pathologies [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eHematologic analysis is essential for diagnosing infections, inflammations, and other disorders in veterinary medicine. For ovine species, changes in WBC subpopulations provide critical insights into health status. Manual microscopic classification, the current gold standard, suffers from inter-observer variability and time inefficiency [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Contemporary developments in artificial intelligence, specifically Convolutional Neural Networks (CNNs), present viable solutions for automated cellular categorization [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. In this study, we develop a CNN-based system for automated classification of five ovine WBC types: neutrophils, eosinophils, basophils, lymphocytes, and monocytes.\u003c/p\u003e"},{"header":"2. LITERATURE REVIEW","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Historical Evolution of WBC Classification (2014\u0026ndash;2024)\u003c/h2\u003e\u003cp\u003eThe field of white blood cell classification has undergone remarkable methodological evolution over the past decade. Early approaches by Putzu et al [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Achieved 93% accuracy using Support Vector Machines with handcrafted shape features for leukemia detection, highlighting the initial focus on specific disease diagnosis rather than comprehensive cell classification. This foundation was subsequently expanded by Şeng\u0026uuml;r et al [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. who integrated shape-based and deep features using LSTM networks, achieving 85.7% accuracy and demonstrating the potential of hybrid methodologies.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eMethodological Evolution and Performance Benchmarking in WBC Classification\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStudy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMethod\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSpecies\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eKey Innovation\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePutzu (2014) [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e93%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSVM\u0026thinsp;+\u0026thinsp;Handcrafted Features\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHuman\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eLeukemia detection\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eŞeng\u0026uuml;r (2019) [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e85.7%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eLSTM\u0026thinsp;+\u0026thinsp;Hybrid Features\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHuman\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eFeature fusion\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSaidani (2023) [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e99%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMulti-fold Pre-processing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHuman\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eData augmentation\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAslan (2023) [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e98.27%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCNN\u0026thinsp;+\u0026thinsp;Feature Selection\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHuman\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eFeature optimization\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAbou Ali (2023) [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSuperior to CNNs (Noisy Data)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eVision Transformer (ViT)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHuman\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eTransformer architecture\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eOur Study (2024)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003e96.72%\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003eCNN\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003eOvine\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003eVeterinary adaptation\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Contemporary Methodological Landscape\u003c/h2\u003e\u003cp\u003eRecent years have witnessed sophisticated architectural innovations. Saidani et al [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] demonstrated the transformative impact of multi-fold pre-processing, achieving 99% accuracy through comprehensive data augmentation. Concurrently, Aslan et al [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e] reached 98.27% accuracy by integrating Ridge feature selection techniques with CNN architectures. The most recent architectural evolution is represented by Abou Ali et al [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e], who introduced Vision Transformers (ViTs) to WBC classification, demonstrating superior performance in handling noisy data. The comprehensive scoping review by Asghar et al [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e] of 136 studies confirmed that CNNs now dominate 54.4% of global research in this domain.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Research Gap Identification and Novel Contributions\u003c/h2\u003e\u003cp\u003eThe comprehensive analysis reveals critical research gaps that our study addresses. Despite extensive research in human hematology, veterinary applications remain severely underexplored. While human-focused studies benefit from large datasets and pre-trained models, veterinary research faces data scarcity and absence of species-specific base models. Our custom CNN architecture demonstrates that excellent performance (96.72%) can be achieved through tailored architectural design, reducing dependency on transfer learning approaches.\u003c/p\u003e\u003c/div\u003e"},{"header":"3. MATERIALS AND METHODS","content":"\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003emodel performance evaluation.\u003c/em\u003e\u003c/p\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Data Collection and Preparation\u003c/h2\u003e\u003cp\u003ePeripheral blood specimens (1 mL volume) were obtained via jugular venipuncture from 36 adult ovine subjects using K2EDTA-containing vacuum tubes [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Giemsa-stained smears were prepared following standard protocols based on established staining methodologies [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. A total of 500 microscopic images were acquired using a 100\u0026times; oil-immersion objective, with an equal distribution of 100 images per WBC type.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Image Preprocessing\u003c/h2\u003e\u003cp\u003eThe preprocessing pipeline consisted of the following sequential steps:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eNoise reduction using median filter with 3\u0026times;3 kernel\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eContrast enhancement through linear stretching\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eGrayscale conversion\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eHistogram equalization\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eResizing to 224\u0026times;224 pixels\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Nuclear Segmentation and Cellular Component Separation\u003c/h2\u003e\u003cp\u003eAdvanced digital image processing techniques were implemented to isolate and preserve nuclear structures.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.4 Image Thresholding and Artifact Removal\u003c/h2\u003e\u003cp\u003eOtsu's optimal thresholding method was employed for pixel classification, followed by size-based filtering to eliminate spurious objects.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.5 CNN Architecture and Training\u003c/h2\u003e\u003cp\u003eA custom CNN architecture was designed using MATLAB 2017a Deep Learning Toolbox:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eInput layer: 224\u0026times;224\u0026times;3\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eThree convolutional blocks (Conv2D\u0026thinsp;+\u0026thinsp;Batch Normalization\u0026thinsp;+\u0026thinsp;MaxPooling)\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eFully connected layer (128 units)\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eOutput layer (5 units with softmax activation)\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eThe model was trained for 50 epochs using Adam optimizer within the MATLAB Deep Learning Toolbox environment, with an initial learning rate set to 0.001. Model performance was evaluated using 5-fold cross-validation to ensure robust generalization.\u003c/p\u003e\u003c/div\u003e"},{"header":"4. RESULTS","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e4.1 Classification Performance\u003c/h2\u003e\u003cp\u003eThe CNN model achieved outstanding performance across all five WBC types as shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, with an overall classification accuracy of 96.72% obtained through 5-fold cross-validation. This overall accuracy represents the\u003c/p\u003e\u003cp\u003emacro-average of individual class accuracies.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eConfusion Matrix of a CNN for Discriminative Classification of Ovine White Blood Cell Types\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWBC Type\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePrecision (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eRecall (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eF1-Score (%)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNeutrophil\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e96.8%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e90.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e94%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e92.1%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMonocytes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e95.8%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e89.1%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e90%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e89.5%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLymphocyte\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e97.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e91.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e96%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e93.6%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEosinophil\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e97.2%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e94.8%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e91%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e92.9%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBasophil\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e96.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e93.6%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e88%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e90.7%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e4.2 Comparison with Manual Classification\u003c/h2\u003e\u003cp\u003eThe automated classification system demonstrated equivalent performance to manual methods for all leukocyte types. Furthermore, the computational approach exhibited enhanced reproducibility with diminished inter-assessment variability.\u003c/p\u003e\u003c/div\u003e"},{"header":"5. DISCUSSION","content":"\u003cp\u003eThe classification accuracy of 96.72% attained in our investigation substantiates exceptional performance when contextualized within the unique morphological challenges of ovine leukocytes. As established in foundational veterinary hematology studies, ovine leukocytes present distinct morphological characteristics not encountered in human samples, including variations in nuclear segmentation patterns and cytoplasmic granularity [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eVeterinary datasets are significantly smaller than their human counterparts, a well-documented challenge in veterinary AI development that often necessitates innovative approaches such as transfer learning from human medical data [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. The absence of pre-trained models for veterinary applications typically requires training from scratch, making our achievement particularly noteworthy.\u003c/p\u003e\u003cp\u003eWhile flow cytometry represents an alternative automated approach, it faces significant challenges in veterinary applications including species-specific reagent limitations and equipment accessibility issues in resource-constrained settings [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. In contrast, our image-based CNN approach provides a more practical solution that leverages standard microscopy equipment already available in most veterinary practices.\u003c/p\u003e"},{"header":"6. CONCLUSION","content":"\u003cp\u003eThis study presents a robust automated system for ovine WBC classification with 96.72% accuracy. The approach offers consistency, efficiency, and objectivity compared to manual methods. While technologies like flow cytometry face implementation barriers in veterinary settings [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e], our image-based approach provides a cost-effective and accessible alternative that maintains high diagnostic accuracy while utilizing existing laboratory infrastructure.\u003c/p\u003e\u003cp\u003eFuture research initiatives will focus on: (1) expanding the dataset to include pathological cases and multiple sheep breeds, (2) implementing stain-invariant models to handle different staining protocols, (3) conducting external validation across multiple veterinary centers, and (4) developing end-to-end systems that integrate cell detection with classification. These enhancements will improve the clinical applicability and robustness of automated hematological analysis in veterinary medicine.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eEthical Statement:\u003c/p\u003e\n\u003cp\u003eThe blood collection procedures in our study have been reviewed and approved by the Research Ethics Committee of Islamic Azad University (Approval ID: IR.IAU.TABRIZ.REC.1402.326). This committee serves as the official ethical oversight body for our research activities, including all procedures involving animals.\u003c/p\u003e\n\u003cp\u003e\u003ch2\u003eCOMPETING INTERESTS\u003c/h2\u003e\u003cp\u003eNo competing interest is declared.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003ch2\u003eETHICAL APPROVAL\u003c/h2\u003e\u003cp\u003e All animal procedures in this study were conducted in accordance with the ethical guidelines of Islamic Azad University, Shabestar Branch.Blood sampling from sheep was performed by experienced veterinarians using standard jugular venipuncture techniques to ensure animal welfare and minimize discomfort.\u003c/p\u003e\u003c/p\u003e\u003ch2\u003eACKNOWLEDGMENTS\u003c/h2\u003e\u003cp\u003eThe authors thank the Islamic Azad University, Shabestar Branch, for providing laboratory facilities and technical support.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAbou Ali M, Dornaika F, Arganda-Carreras I (2023) White Blood Cell Classification: Convolutional Neural Network (CNN) and Vision Transformer (ViT) under Medical Microscope, vol 242. Computer Methods and Programs in Biomedicine, p 107839\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAsghar R, Kumar S, Hynds P, Shaukat A (2023) Classification of White Blood Cells Using Machine and Deep Learning Models: A Systematic Review. Artif Intell Med 146:102687\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAslan E, \u0026Ouml;z\u0026uuml;pak Y (2024) Classification of Blood Cells with Convolutional Neural Network Model, vol 62. Medical \u0026amp; Biological Engineering \u0026amp; Computing, pp 345\u0026ndash;356\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAtkins CG, Buckley K, Blades MW, Turner RFB (2017) Raman Spectroscopy of Blood and Blood Components. Appl Spectrosc 71(5):767\u0026ndash;793\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBaruah DK, Boruah K (2025) Comparative Study of CNN-ML Models for Detection and Classification of Canine Babesia in Blood Cell Image. Int J Next-Generation Comput 16(1):78\u0026ndash;92\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBollig N, Nichelason A (2021) Transfer learning: leveraging human data to build smarter veterinary AI. J Veterinary Med Inf 28(2):45\u0026ndash;52\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDunning K, Safo AO (2011) The ultimate Wright-Giemsa stain: 60 years in the making. Biotech Histochem 86(2):69\u0026ndash;75\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHegde RB, Prasad K, Hebbar H, Singh B (2019) Feature extraction using traditional image processing and convolutional neural network methods to classify white blood cells: a study, vol 42. Australasian Physical \u0026amp; Engineering Sciences in Medicine, pp 50\u0026ndash;60\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHunka J, Riley JT, Debes G (2020) Approaches to overcome flow cytometry limitations in the analysis of cells from veterinary relevant species. BMC Vet Res 16(1):239\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJones DC, Krebs JS (1972) Hematologic characteristics of sheep. Am J Vet Res 33(5):1025\u0026ndash;1031\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKareem SW (2021) An Evaluation Algorithms for Classifying Leukocytes Images. J Med Imaging Health Inf 11(4):1123\u0026ndash;1130\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLayssol-Lamour C, Granat F, Sahal A, Braun J, Trumel C, Bourg\u0026egrave;s-Abella N (2022) Improving the Quality of EDTA-treated Blood Specimens from Mice. J Am Assoc Lab Anim Sci 61(3):235\u0026ndash;243\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMishra AK, Jatav VK, Bairwa K (2023) A Study of Blood Cells, Work and Function. Int J Hematol Res 9(2):45\u0026ndash;58\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePutzu L, Caocci G, Di Ruberto C (2014) Leucocyte classification for leukaemia detection using image processing techniques. Artif Intell Med 62(3):179\u0026ndash;191\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRamadevi P, Vidyulatha G, Mahesh L (2021) Blood Cell Classification using Deep Learning CNN Model. Int J Adv Comput Sci Appl 12(7):123\u0026ndash;130\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSaidani O, Umer M, Alturki NM, Alshardan A, Kiran M, Alsubai S, Kim T, Ashraf I (2024) White blood cells classification using multi-fold pre-processing and optimized CNN model. Sci Rep 14:15642\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eŞeng\u0026uuml;r A, Akbulut Y, Budak \u0026Uuml;, C\u0026ouml;mert Z (2019) White Blood Cell Classification Based on Shape and Deep Features. In 2019 International Artificial Intelligence and Data Processing Symposium (IDAP) (pp. 1\u0026ndash;5). IEEE\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Islamic Azad University Shabestar","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Ovine Leukocyte, WBC Classification, Convolutional Neural Network, Image Processing, Veterinary Hematology","lastPublishedDoi":"10.21203/rs.3.rs-7790796/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7790796/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eThe differential classification of white blood cells (WBCs) is crucial for diagnosing various health conditions in veterinary medicine. Traditional manual classification is time-consuming and subjective.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eThis study presents an automated system for classifying five types of ovine WBCs using image processing and Convolutional Neural Networks (CNN). Blood samples from 36 adult sheep were used to create 500 microscopic images (100 per cell type). A comprehensive image processing pipeline was implemented in MATLAB and a custom CNN architecture was designed for five-class classification.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eThe CNN model achieved an overall classification accuracy of 96.72% through 5-fold cross-validation, demonstrating robust performance across all five leukocyte types.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eThe proposed system provides a reliable, efficient, and accurate tool for automated WBC classification in sheep, with potential applications in veterinary diagnostics.\u003c/p\u003e","manuscriptTitle":"Automated classification of five types of ovine white blood cells using Convolutional Neural Network and image processing","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-12 14:31:44","doi":"10.21203/rs.3.rs-7790796/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"000f84eb-4e55-4217-8693-f3896bb1aa5a","owner":[],"postedDate":"October 12th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":55829870,"name":"Bioinformatics"},{"id":55829871,"name":"Animal Science"},{"id":55829872,"name":"Artificial Intelligence and Machine Learning"},{"id":55829873,"name":"Biotechnology and Bioengineering"}],"tags":[],"updatedAt":"2025-10-12T14:31:44+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-12 14:31:44","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7790796","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7790796","identity":"rs-7790796","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}