泌乳牛乳房炎的早期檢測對乳房炎的防治是相當重要,本研究目的在使用近紅外線光譜分析技術,以不同的光譜數學前處理方法,利用迴歸分析方法獲得乳酸去氫酵素含量的乳房炎光譜檢量線,並評估乳房炎光譜檢量線的檢測效率;研究採集泌乳牛各分房前乳,分別量測其體細胞數及乳酸去氫酵素,並進行近紅外線光譜掃瞄。實驗分析結果可知,經過微分處理後的近紅外線光譜校正線可得到的較小的校正偏差SEC及較高的決定係數R2,修正部分最小平方迴歸得到乳酸去氫酵素的乳房炎光譜檢量線其校正標準偏差SEC為16.532,決定係數R2為0.859;此乳房炎光譜檢量線對健康分房的特定度為81.6%;對感染分房的敏感度為85.7%。
Detecting mastitis of dairy cows at the early stage is a very important task for disease prevention and treatment. A near-infrared (NIR) spectral calibration algorithm for eveluation of mastitis detective efficacy is developed by means of various spectral mathematical pretreatment processing. The algorithm was used for rapid content detection of lactate dehydronase in quarter foremilk of dairy cows. A series of analyses on somatic cell counts (SCC), lactate dehydronase (LDH), and near-infrared scanning were conducted on the same quarter foremilk samples. The experimental results revealed that the calibration algorithm based on mathematical derivatives methods could reduce the standard error calibration (SEC) and increase the coefficient of determination R^2 effectively. A LDH mastitis spectra calibration equation was hence derived from modified partial least squares (MPLS) regression, which resulted in 16.532 of SEC and 0.859 of R^2. The specificity of a healthy quarter and sensitivity of an infected quarter from the mastitis spectral calibration equation are 81.6% and 85.7%, respectively.