The present application provides a method, apparatus, device and medium for optimizing thresholds based on K-edge imaging. The method comprises: pre-scanning, by using a photon counting detector, an object to be detected according to a preset threshold group, to obtain first detection data; determining, according to the first detection data, line integral values of decomposition coefficients of individual basis functions in a basis function combination of each ray, wherein each ray has a fixed basis function combination; selecting, according to the line integral values of the decomposition coefficients of individual basis functions in the basis function combination of the each ray, rays of which a line integral value of a contrast agent material basis function in the fixed basis function combination is greater than a preset line integration threshold; inputting detection data corresponding to each of the selected rays to a pre-trained neural network model, to obtain a local optimal threshold group of each of the selected rays; and determining, based on local optimal threshold groups, a global optimal threshold group for the K-edge imaging of the photon counting detector.