Machine Learning Algorithms for Peripheral Blood Cell Classification - A Hemovision Project Experience

Aug 3, 2026

Upload by Finn

Research presents machine learning approaches for classifying nucleated peripheral blood cells using a Hemovision project workflow. A ResNet18 convolutional neural network performs image preprocessing and replaces dense layers, while a Support Vector Machine (SVM) classifier is used for output. Training and testing rely on images collected from different datasets to evaluate generalization. Results report an accuracy and F1-Score of 99.96%, supporting integration of machine learning into educational practice and diagnostic support processes.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Introduction

Background and clinical relevance of CBC and blood smear analysis

Role of automated analyzers vs. need for morphological expertise

Machine learning and deep learning methods for image-based classification

Candidate algorithms and available blood cell databases

Revista de Informtica Terica e Aplicada-RITA-ISSN 2175-2745 Vol. 30, Num. 02 (2023) 89-100 RESEARCH ARTICLE Machine Learning Algorithms for Peripheral Blood Cell Classification-A Hemovision Project Experience Algoritmos de Aprendizado de Mquina para Classificac¸ o de Clulas do Sangue Perifrico-Uma Experincia do Projeto Hemovision Mariana Dourado X. S. Santos 1 *, William Laus Bertemes2 , Iaan Mesquita de Souza 1 , Mateus Henrique B. Andrades 1 , Vinicius Sebba Patto 1 |
Abstract: This research explores the use of machine learning algorithms to classify nucleated peripheral blood cells. The ResNet18 convolutional neural network was used to pre-process the images and replace the dense layers; and for the output, the Support Vector Machine (SVM) classifier was chosen. Images from different datasets were used for training and testing the model. Thus, the developed model achieved an accuracy and F1-Score of 99.96% . In face of the obtained results, it was found that machine learning algorithms can be satisfactorily integrated into educational and diagnostic support processes.
Keywords: Support Vector Machine—Convolutional Neural Networks—White Blood Cells—Machine Learning
Resumo: Este trabalho trata do uso de algoritmos de aprendizado de mquina para classificac¸o de clulas nucleadas do sangue perifr˜ico. Foi utilizada a rede neural convolucional ResNet18 para o pr-processamento
das imagens e em substituic¸ao s camadas densas; e para a sa´ıda foi escolhido o classificador Support Vector Machine (SVM) . Foram usadas imagens de diferentes datasets para o treino e teste do modelo. Assim, o modelodesenvolvido alcanc¸ou uma acurcia e F1-Score de 99 .96% . Diante dos resultados encontrados, constatou-se que os algoritmos de aprendizado de mquina podem ser integrados de forma satisfatria aos processos educacionais e de apoio ao diagnstico.
Palavras-Chave: Mquina de Vetor de Suporte—Redes Neurais Convolucionais—Glbulos Brancos—Aprendizado de Mquina | | --- | | 1 Instituto de Informa´ tica, Universidade Federal de Goia´ s (UFG), Goinia - Goia´ s, Brazil
2 Hospital das Cl´ınicas, Universidade Federal de Goia´ s (UFG), Goinia - Goia´ s, Brazil
*[Corresponding author](Corresponding author: marianadximenes@gmail.com)[: marianadximenes@gmail.com](Corresponding author: marianadximenes@gmail.com)
DOI: [http://dx.doi.org/10.22456/2175-2745.130669](http://dx.doi.org/10.22456/2175-2745.130669) • Received: 06/03/2023 • Accepted: 22/06/2023
CC BY-NC-ND 4 .0 - This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4 . 0 International License. | 1. Introduction The complete blood count (CBC) is the most requested complementary exam in the world; it is an integral part of health screening and an almost indispensable adjunct in the diagnosis and follow-up of chronic diseases in general, infectious diseases, medical emergencies, and the follow-up of chemotherapy and radiotherapy, having a broad relationship with the other clinical pathology exams [1] . With the development of sophisticated automated blood cell analyzers to perform the CBC, the number of samples requiring blood distension for morphological evaluation has greatly decreased and in many clinical settings is around 10% to 15% of routine samples. Nevertheless, blood distention remains a crucial diagnostic aid [2] . In order to obtain the most clinically relevant information from the morphological analysis of a blood distention, the procedure should be performed by a person very experienced in hematological microscopy, be it a clinical pathologist ora medical hematologist. Given that, knowing the unusual changes in the peripheral blood and its morphology allows the analyst to make accurate diagnoses and suggest, in a targeted and cost-effective manner, the performance of further complementary tests [2] [3] . However, often this experience is developed over several years and after experiencing different clinical case

FAQs

What is the document’s main goal?

It explores how machine learning algorithms can classify nucleated peripheral blood cells using a Hemovision project experience.

Which model components are used for preprocessing and classification?

ResNet18 is used for image preprocessing and architectural replacement of dense layers, and an SVM classifier is used for the output.

How is the model evaluated, and what performance is achieved?

Images from different datasets are used for training and testing, achieving an accuracy and F1-Score of 99.96%.

Aug 3, 2026

Upload by Finn