Franca Thiago, Lacerda Miller, Calvani Camila, Arruda Kelvy, Maranni Ana, Nicolodelli Gustavo, Karthikeyan Sivakumaran, Marangoni Bruno, Nascimento Carlos, Cena Cicero
Optics and Photonic Lab (SISFOTON-UFMS), UFMS-Universidade Federal de Mato Grosso do Sul, Av. Costa e Silva s/n, Campo Grande, MS 79070-900, Brazil.
UFSC-Universidade Federal de Santa Catarina, Rua Eng Argonomico Andrei Cristian s/n, Florinópolis, SC 88040-900, Brazil.
ACS Omega. 2025 May 24;10(22):22952-22959. doi: 10.1021/acsomega.5c00504. eCollection 2025 Jun 10.
Diagnosing bovine brucellosis is a major challenge due to its significant economic impact, causing losses in meat and dairy production and its potential to transmit to humans. In Brazil, disease control relies on diagnosis, animal culling, and vaccination. However, existing diagnostic tests, despite their quality, are time-consuming and prone to false positives and negatives, complicating effective control. There is a critical need for a low-cost, fast, and accurate diagnostic test for large-scale use. Spectroscopy techniques combined with machine learning show great promise for improving diagnostic tests. Here, we explore the potential use of FTIR (Fourier transform infrared) spectroscopy and machine learning algorithms to provide a rapid, accurate, and cost-effective diagnostic method for Brucella abortus. This study explored the use of FTIR spectroscopy on bovine blood serum in liquid and dried forms to develop a new photodiagnosis method. Eighty bovine blood serum samples (40 infected and 40 control animals) were analyzed. Initially, the FTIR data were pretreated using the standard normal deviate method to remove baseline deviations. Principal component analysis was then applied to observe clustering tendencies, and the further selection of principal components improved clustering. Using support vector machine algorithms, the predictive models achieved overall accuracies of 95.8% for dried samples and 91.7% for liquid samples. This new methodology delivers results in about 5 min, compared to the 48 h required for standard diagnostic methods. These findings demonstrate the viability of this approach for diagnosing bovine brucellosis, potentially enhancing disease control programs in Brazil and beyond.
诊断牛布鲁氏菌病是一项重大挑战，因为它会产生重大经济影响，导致肉类和奶制品生产损失，并且有可能传染给人类。在巴西，疾病控制依赖于诊断、扑杀动物和接种疫苗。然而，现有的诊断测试尽管质量不错，但耗时较长，且容易出现假阳性和假阴性结果，给有效控制带来了困难。迫切需要一种低成本、快速且准确的大规模诊断测试。光谱技术与机器学习相结合，在改进诊断测试方面显示出巨大潜力。在此，我们探索了傅里叶变换红外光谱（FTIR）和机器学习算法在提供一种快速、准确且经济高效的布鲁氏菌诊断方法方面的潜在用途。本研究探讨了使用FTIR光谱对液态和干燥形式的牛血清进行分析，以开发一种新的光诊断方法。分析了80份牛血清样本（40份来自感染动物，40份来自对照动物）。最初，使用标准正态变量法对FTIR数据进行预处理，以消除基线偏差。然后应用主成分分析来观察聚类趋势，进一步选择主成分改善了聚类效果。使用支持向量机算法，预测模型对干燥样本的总体准确率达到了95.8%，对液态样本的总体准确率达到了91.7%。这种新方法在大约5分钟内就能得出结果，而标准诊断方法则需要48小时。这些发现证明了这种方法在诊断牛布鲁氏菌病方面的可行性，有可能加强巴西及其他地区的疾病控制计划。
