Aprimoramento da Classificação de Linfócitos e Monócitos em Imagens Médicas: O Impacto do CLAHE em Redes Neurais Convolucionais

The human immune system plays a critical role in defending againstinfections and diseases, with white blood cells (WBCs) being pivotalin these processes. Automated classification of agranulocytecells, specifically lymphocytes, and monocytes, is essential for accuratediagnostics and treatment monitoring in hematology andoncology. This study evaluates the performance of a convolutionalneural network (CNN) model, previously proposed for WBC classification,on public datasets, with and without the use of ContrastLimited Adaptive Histogram Equalization (CLAHE) for image preprocessing.The results show that CLAHE improved classificationmetrics, achieving up to 82.16% test accuracy on the Paul Mooneydataset and maintaining a high test accuracy of 98.72% on the UncleSamulus dataset. Metrics such as precision, recall, and F1-score alsoexhibited notable improvements, reaching up to 98% for lymphocytesand monocytes in the best-performing dataset. These findingshighlight CLAHE’s potential to enhance CNN-based classificationunder varying image conditions.