The invention relates to the technical field of medical image processing, in particular to a focus tracking method under a digestive endoscope based on sequential feature learning, which comprises thefollowing steps: collecting lesion video clips as training samples, and constructing and training a tracking model based on a convolutional neural network, a long short-term memory network and an optical flow vector diagram; acquiring network model parameters, acquiring a digestive endoscopy real-time examination video, deframing the digestive endoscopy real-time examination video into pictures,calculating optical flow vector diagrams of two adjacent frames of images, loading a network structure and model parameters based on a convolutional neural network, a long short-term memory network and the optical flow vector diagrams, and calculating the area and position of a focus in real time. By means of the method, the area and the position of the focus can be tracked in real time in the digestive endoscopy process for endoscopy doctors to refer to, and the situation that the focus area is lost due to illumination, angles, shielding and other reasons in the examination process can be effectively prevented. The detection and tracking capacity of the focus under the digestive endoscopy can be improved, and the examination quality of the digestive endoscopy is effectively improved.