Inovação no Diagnóstico de Brucelose e Tuberculose Bovina: Métodos de Fotodiagnóstico Baseados em Espectroscopia FTIR e Machine Learning

• Ano de defesa: 2025
• Autor(a) principal: THIAGO FRANCA DA SILVA
• Orientador(a): Cicero Rafael Cena da Silva
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Fundação Universidade Federal de Mato Grosso do Sul
• País: Brasil
• Palavras-chave em Português: Brucelose; Tuberculose Bovina; Espectroscopia FTIR
• Link de acesso: https://repositorio.ufms.br/handle/123456789/11105

Bovine brucellosis and tuberculosis are zoonotic diseases with significant impacts on public health, animal health, and the economy. Caused by bacteria of the genera Brucella and Mycobacterium, respectively, these diseases are primarily transmitted to humans through contaminated animal-derived products or direct contact with infected animals. This study evaluated the use of Fourier Transform Infrared (FTIR) spectroscopy combined with machine learning as a screening tool for diagnosing these infections using bovine blood serum samples. The study investigated the influence of sample preparation conditions on classification performance, divided into three stages. Initially, samples from a Control group and a Brucellosis group (animals confirmed to be infected with Brucella abortus) were compared using two approaches: oven-dried serum, a well-established method in the literature, and liquid serum with a deionized water background, an innovative, practical, and rapid alternative. Subsequently, with liquid serum samples, binary classifications were conducted between Control vs. Tuberculosis (animals confirmed to be infected with Mycobacterium bovis) and Control vs. Brucellosis, enabling model validation and optimization for multiclass classification. Finally, multiclass classification was applied to simultaneously distinguish among the three groups (Control vs. Brucellosis vs. Tuberculosis). The results demonstrated that liquid serum outperformed dried serum, achieving 100% accuracy and sensitivity in diagnosing brucellosis, surpassing conventional methods. Although the accuracy for tuberculosis was 83.3%, the multiclass approach achieved 90.5% accuracy and up to 100% sensitivity, underscoring the method's effectiveness in differentiating between control and infected animals. The analysis revealed the joint contribution of vibrational modes from molecules belonging to different groups, such as lipids, proteins, and carbohydrates. The combination of FTIR spectroscopy with machine learning, using liquid blood serum with a deionized water background, proved to be an innovative, rapid, and efficient method, eliminating complex sample preparation steps. This approach holds potential for on-site diagnostics and the control of these zoonoses, reducing their spread among animals and humans. Keywords: Bovine brucellosis, Bovine tuberculosis, FTIR spectroscopy, Machine learning, Onsite diagnosis.