Least squares SVM combined with near infrared spectroscopy for diagnosing endometrial carcinoma

Abstract

The feasibility of early diagnosis of endometrial carcinoma was studied by least squares support vector machines (LS-SVM) that classified near infrared (NIR) spectra of tissues. NIR spectra of 77 specimens of endometrium were collected. The spectra were pretreated by the 1st derivative Savitzky-Golay and direct orthogonal signal correction (DOSC) methods to improve the signal-to-noise ratio and remove the influences of background and baseline. The effects of modeling parameters were investigated using grid searching technique and bootstrapped Latin-partition methods. The model was optimized with the destination function of the average RMSE of bootstrapped Latinpartition cross validation. The optimal model of the LS-SVM successfully classified all the samples of the test set into three groups. The proposed procedure was proven to be rapid and convenient, which is suitable to be developed as a non-invasive diagnosis method for cancer tissue.