In a groundbreaking development, researchers from Bangladesh Agricultural University (BAU) have created a machine learning-based tool that can predict the likelihood of brucellosis in cattle—eliminating the need for costly laboratory tests and offering a faster, data-driven solution to livestock disease management.
The project was led by Professor Dr. Md. Siddiqur Rahman, Head of the Department of Medicine at BAU's Faculty of Veterinary Science, in collaboration with PhD researcher Col. (Retd.) SM Azizul Karim Hossaini. The study, supported by Germany’s Friedrich Loeffler Institute and King Faisal University in Saudi Arabia, was published in the ‘Asian Journal of Agriculture and Biology’ on May 4.
Using reproductive and health data from two major dairy farms in Savar, the researchers collected 460 cattle samples and performed the Rose Bengal Test to confirm brucellosis. To process the data, they used the Synthetic Minority Over-sampling Technique (SMOTE) and tested five machine learning algorithms. The Multilayer Perceptron (MLP) and Weka Deep Learning 4J algorithms achieved the highest accuracy rates of 93.59 percent and 93.94 percent, respectively.
Professor Rahman emphasized that this is the first use of machine learning in Bangladesh for predicting brucellosis in cattle. The tool allows farmers to assess infection risk by simply inputting farm management data—enabling early action and biosecurity measures.
Brucellosis, caused by the ‘Brucella’ bacteria, is a zoonotic disease that can severely impact animal fertility and milk production, and pose reproductive health risks in humans. The researchers also found that a combination of oxytetracycline, streptomycin, and benzylpenicillin offers effective treatment, while killed vaccines provide safer protection compared to live variants.
Col. Karim added that the innovation will reduce veterinary costs, improve herd health, and pave the way for future vaccine breakthroughs—marking a significant advancement in Bangladesh’s livestock sector.
