PDF(Pubmed)
Abstract：
Subclinical mastitis is a common and economically significant disease that affects dairy sheep production. Thermal imaging presents a promising avenue for non-invasive detection, but existing methodologies often rely on simplistic temperature differentials, potentially leading to inaccurate assessments. This study proposes an advanced algorithmic approach integrating thermal imaging processing with statistical texture analysis and t-distributed stochastic neighbor embedding (t-SNE). Our method achieves a high classification accuracy of 84% using the support vector machines (SVM) algorithm. Furthermore, we introduce another commonly employed evaluation metric, correlating thermal images with commercial California mastitis test (CMT) results after establishing threshold conditions on statistical features, yielding a sensitivity (the true positive rate) of 80% and a specificity (the true negative rate) of 92.5%. The evaluation metrics underscore the efficacy of our approach in detecting subclinical mastitis in dairy sheep, offering a robust tool for improved management practices.
摘要：
亚临床乳腺炎是影响乳羊生产的常见且具有经济意义的疾病。热成像为非侵入性检测提供了一个有希望的途径，但是现有的方法通常依赖于简单的温差，可能导致不准确的评估。本研究提出了一种先进的算法方法，将热成像处理与统计纹理分析和t分布随机邻居嵌入（t-SNE）集成在一起。我们的方法使用支持向量机（SVM）算法实现了84％的高分类精度。此外，我们介绍另一种常用的评估指标，在建立统计特征的阈值条件后，将热图像与商业加利福尼亚乳腺炎测试（CMT）结果相关联，产生80%的敏感性(真阳性率)和92.5%的特异性(真阴性率)。评估指标强调了我们的方法在检测奶牛亚临床乳腺炎中的功效，提供一个强大的工具来改进管理实践。
