The invention provides an image-based computer-aided analytical method for performing non-invasive monitoring to the degree of tumor-infiltrated surrounding tissue and the method can reflect the degree of tumor-infiltrated surrounding tissue. Therefore, the invention provides a safe and non-invasive mean of monitoring, screening and diagnosing tumor diseases. The method comprises the following steps: 1) obtaining the image of a special position, selecting an interesting area; 2) dividing a ROI area into a plurality of 'reference units'; 3) calculating the textural characteristic of the 'neighboring areas' surrounding the 'reference units'; 4) using the textural characteristic value obtained in the previous step to form a characteristic vector, sending the characteristic vector to a classifier for detection, obtaining the attribute of the 'neighboring areas'; 5) calculating the score of the 'reference units' according to the attribute of the 'neighboring areas' surrounding the 'reference units', obtaining the probability distribution map of the ROI area; 6) overlapping the probability distribution map with the original image to obtain the boundary of tumor tissue and normal tissue; and 7) correcting the boundary of benign and malignant tissue according to the growth characteristics of tumor so as to finally determine the degree of tumor-infiltrated surrounding tissue.