A method, device and system for automatic segmentation of multi-target organs at risk based on deep learning

Abstract

The invention discloses a method, device and system for automatically segmenting multi-target organs at risk based on deep learning. The method includes: receiving an input image of a patient; converting the input image of the patient into JPEG format data; inputting the JPEG format data According to the Overfeat location detection network trained by the physicist's manual segmentation results, the region of interest including multi-target organs at risk is automatically selected; the automatically selected region of interest is input into the FCN initialization segmentation network for contour inference; the initial boundary contour obtained by contour inference Coordinate with the received artificially marked boundary, map to the input image, extract the DAISY feature, and obtain the DAISY feature image; input the DAISY feature image into the deep belief network trained according to the manual segmentation results of the physicist, and obtain the precise segmentation boundary of the organ at risk, namely Split results.

Description

Technical field [0001] This disclosure is a technical field that endangers organ segmentation, involving a multi-objective risk of deep learning, and automatic segmentation methods, devices, and systems. Background technique [0002] The statement of this section is merely the background technology information related to the present disclosure, which is not necessarily constituted in prior art. [0003] Pituitoma is a group of tumors that occur from pituitary front leaf and post leaf and craniopharyngeal tuberculosis, the most important harm of pituitary tumors on the human body is the structure of the butterfly, such as the view, optic nerves, sponge sinus, brainstem, Brain bottom artery, etc., causing a serious obstacle to the corresponding function. Among them, the view god cross, the optic nerve and brainstem are most common by pituitary tumors. Pituitous tumor compression, levity, leading to patient vision, vision change, blindness and other obstacles; vertebral tumor pressi...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the shortcomings of the prior art that require manual selection of features and poor algorithm robustness, one or more embodiments of the present disclosure provide a method, device and system for automatic multi-target organ-at-risk segmentation based on deep learning, Receive the MR image of the patient, use the advanced deep learning neural network to automatically locate and detect the position of the organ at risk, and then use the automatic segmentation network to learn the manual segmentation results of the physicist for initial segmentation, and then use the precise segmentation to obtain the position and contour information of the organ at risk , can automatically segment the optic chiasm, optic nerve and brainstem at the same time, and send the output results to the physicist's surgery planning system to assist gamma knife radiation therapy for pituitary tumors

Images