The invention relates to screening and calibration of information in the technical field of information, and discloses a method for performing automatic segmentation based on a focus tomography scan,which comprises the following steps: firstly, establishing a data set: acquiring a plurality of section scan images of a target part; preprocessing and labeling the acquired section scan image, and labeling a target tissue and a non-target tissue of the section scan image for distinguishing, so as to obtain a plurality of labeled samples; and storing the labeled samples to obtain a data set; and secondly, training a model: establishing a convolutional neural network model, inputting the labeled samples for training, and finally outputting the trained convolutional neural network model; and finally, performing segmentation: preprocessing the tomoscan image, and inputting the preprocessed tomoscan image into the trained convolutional neural deep learning model for segmentation. The method isapplied to particle implantation, information identification and acquisition of the tomography images can be rapidly achieved after a model is trained, the required tomography images are obtained through segmentation, and the accuracy and efficiency of particle implantation can be improved conveniently.