Estudo Comparativo de Algoritmos de Classificação de Imagens na Identificação de Pneumonia em Raios-x Pulmonares

• Jennifer Mayara de Paiva Goberski Universidade Positivo
• João Vitor Maciel de Brito Universidade Positivo
• Marcel Antunes Raposo Universidade Positivo
• Vinicius Dionizio Patrocinio Universidade Positivo
• Leonardo Gomes Tavares Universidade Positivo

DOI: https://doi.org/10.5753/sibgrapi.est.2024.31666

Resumo

Este trabalho apresenta uma análise comparativa de diferentes modelos de técnicas de aprendizado de máquina para a classificação binária de imagens de raio-x pulmonares a fim de de identificar a presença ou ausência de pneumonia. Utilizando uma base de dados pré-processada, que incluiu normalização e balanceamento de classes, foram aplicadas técnicas de data augmentation e pesos de classe para otimizar o treinamento dos modelos. Entre os modelos clássicos avaliados, como Regressão Logística, Árvore de Decisão, SVM, MLP e Random Forest, a SVM destacou-se com a melhor performance, obtendo uma pontuação média de F1 score de 0,96. No entanto, os modelos de aprendizado profundo, particularmente as Redes Neurais Convolucionais (CNNs), como ResNet e EfficientNet, superaram significativamente os modelos tradicionais, alcançando F1 scores de até 0,98. A ResNet foi identificada como a arquitetura mais eficaz, graças à sua capacidade de capturar características complexas nas imagens de raio-x. Os resultados sugerem que o uso desses modelos pode aprimorar a precisão diagnóstica, oferecendo suporte valioso para profissionais de saúde na detecção precoce de doenças pulmonares.

Palavras-chave: Aprendizado de máquina, aprendizado profundo, pneumonia, raio-x, diagnóstico

Referências

Alsharif, Roaa and Al-Issa, Yazan and Alqudah, Ali Mohammad and Qasmieh, Isam Abu and Mustafa, Wan Azani and Alquran, Hiam, ”PneumoniaNet: Automated Detection and Classification of Pediatric Pneumonia Using Chest X-ray Images and CNN Approach”. Electronics, v. 10, n. 23, p. 2949, 2021.

Buda, Mateusz and Maki, Atsuto and Mazurowski, Maciej A, ”A systematic study of the class imbalance problem in convolutional neural networks”. Neural networks, v. 106, p. 249-259, 2018. Elsevier.

Chee, Elyssa and Huang, Kathryn and Haggie, Stuart and Britton, Philip N, ”Systematic review of clinical practice guidelines on the management of community acquired pneumonia in children”. Paediatric Respiratory Reviews, v. 42, p. 59-68, 2022. Elsevier.

Corrêa, Ricardo de Amorim and Costa, Andre Nathan and Lundgren, Fernando and Michelin, Lessandra and Figueiredo, Mara Rúbia and Holanda, Marcelo and Gomes, Mauro and Teixeira, Paulo José Zimermann and Martins, Ricardo and Silva, Rodney and others, ”2018 recommendations for the management of community acquired pneumonia”. Jornal Brasileiro de Pneumologia, v. 44, n. 05, p. 405-423, 2018. SciELO Brasil.

Dovganich, A. A. and Khvostikov, A. V. and Krylov, A. S. and Parolina, L. E., ”Automatic quality control in lung X-ray imaging with deep learning”. Computational Mathematics and Modeling, v. 32, p. 276-285, 2021.

Ferreira, José Raniery and Cardenas, Diego Armando Cardona and Moreno, Ramon Alfredo and de Sá Rebelo, Marina de Fátima and Krieger, José Eduardo and Gutierrez, Marco Antonio, ”Multi-view ensemble convolutional neural network to improve classification of pneumonia in low contrast chest x-ray images”. In: 2020 42nd annual international conference of the IEEE engineering in Medicine & Biology Society (EMBC). IEEE, p. 1238-1241, 2020.

Hasan, Md Jahid and Alom, Md Shahin and Ali, Md Shikhar, ”Deep learning based detection and segmentation of COVID-19 & pneumonia on chest X-ray image”. In: 2021 International Conference on Information and Communication Technology for Sustainable Development (ICICT4SD). IEEE, p. 210-214, 2021.

Ikechukwu, A Victor and Murali, S and Deepu, R and Shivamurthy, RC, ”ResNet-50 vs VGG-19 vs training from scratch: A comparative analysis of the segmentation and classification of Pneumonia from chest X-ray images”. Global Transitions Proceedings, v. 2, n. 2, p. 375-381, 2021. Elsevier.

Keras, ”Keras 3 API documentation: Keras Applications”. 2024. [link]. Acessado em 13/07/2024.

Mujahid, Muhammad and Rustam, Furqan and Álvarez, Roberto and Luis Vidal Mazón, Juan and Díez, Isabel de la Torre and Ashraf, Imran, ’Pneumonia classification from X-ray images with inception-V3 and convolutional neural network”. Diagnostics, v. 12, n. 5, p. 1280, 2022. MDPI.

Sharma, Harsh and Jain, Jai Sethia and Bansal, Priti and Gupta, Sumit, ”Feature extraction and classification of chest x-ray images using cnn to detect pneumonia”. In: 2020 10th international conference on cloud computing, data science & engineering (Confluence). IEEE, p. 227-231, 2020.

Ouerhani, Amira and Boulares, Souhaila and Mahjoubi, Halima, ”Automated Detection of Pediatric Pneumonia from Chest X-Ray Images Using Deep Learning Models”. In: 2023 IEEE Afro-Mediterranean Conference on Artificial Intelligence (AMCAI). IEEE, p. 1-7, 2023.

Stephen, Okeke and Sain, Mangal and Maduh, Uchenna Joseph and Jeong, Do-Un, ”An efficient deep learning approach to pneumonia classification in healthcare”. Journal of healthcare engineering, v. 2019, n. 1, p. 4180949, 2019. Wiley Online Library.

Xu, Wanni and Fu, You-Lei and Zhu, Dogmei, ”ResNet and its application to medical image processing: Research progress and challenges”. Computer Methods and Programs in Biomedicine, v. 240, p. 107660, 2023.

Yun, Ki Wook and Wallihan, Rebecca and Juergensen, Alexis and Mejias, Asuncion and Ramilo, Octavio, ”Community-acquired pneumonia in children: myths and facts”. American journal of perinatology, v. 36, n. S 02, p. S54-S57, 2019. Thieme Medical Publishers.