Hybrid 2D-3D medical image real-time segmentation and intraoperative navigation method and related device

Abstract

The invention discloses a hybrid 2D-3D medical image real-time segmentation and intraoperative navigation method and a related device. The method comprises the following steps: acquiring a 2D slice sequence of an intraoperative target organ and preprocessing the 2D slice sequence; adjacent 2D slices are stacked in the Z-axis direction according to the spatial sequence to form 3D volume data, post-processing is carried out, and 3D feature representation is obtained; inputting the 3D feature representation into a medical image segmentation module, performing processing based on a deep learning algorithm to obtain a segmentation mask and confidence, and reminding a doctor to perform manual segmentation when the confidence is lower than 85%; registering the segmentation mask with the intraoperative medical image data by adopting a rigid or non-rigid registration algorithm, and displaying on an interface of the surgical navigation equipment; the large model intelligent agent is used for coordinating and controlling the working process and data transmission of each step, so that the real-time navigation in the operation is realized. According to the invention, image segmentation and navigation can be completed in real time in an operation, and the operation precision and efficiency are improved.

Images