Fig 2. Early screening for four representative cancers by SERS-AICS. (A) The SERS-AICS method could also effectively distinguish common diseases from early cancers in stage I and II with high accuracy. (B) ROC curves with covariance matrices-assisted SVM model for distinguishing 45 common disease patients from 33 early stage of cancer patients about lung. (C) ROC curves with covariance matrices-assisted SVM model for distinguishing 42 common disease patients from 32 early stage of cancer patients about colorectum. (D) ROC curves with covariance matrices-assisted SVM model for distinguishing 39 common disease patients from 36 early stage of cancer patients about gastric. (E) ROC curves with covariance matrices-assisted SVM model for distinguishing 33 common disease patients from 32 early stage of cancer patients about liver. (F) The accuracy, sensitivity and specificity of different common disease/early stage of cancer.

Fig 3. Construction of cancer-related database at bond level by SERS-AICS. (A) The original spectral data of 1465 dimensions of serum with no obvious specificity can obtain 50 valid dimensions with the best specificity after the correlation between dimensions is judged by the covariance matrix, which related to the molecular bond energy information in serum associated with cancer or disease. (B) Heatmap of the covariance matrix formed on 30 true dimensions between 600-623.77051 cm-1 using peak data for lung cancer and healthy controls, the dimension at 600 cm-1 and 618.03276 cm-1 showed the minimal correlation. (C) List of common valid dimensions of different cancers compared with normal control group or common disease.