A TEE heart ultrasound real-time segmentation method based on Unet network
The TEE cardiac ultrasound real-time segmentation method based on Unet network solves the problems of resource imbalance, low efficiency and unstable imaging in traditional ultrasound diagnosis. It realizes automatic segmentation of lesion areas and parameter measurement, improves diagnostic accuracy and efficiency, and promotes the fairness and inclusiveness of medical resources.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING LINJIN SPACE AIRCRAFT SYST ENG INST
- Filing Date
- 2026-06-02
- Publication Date
- 2026-07-24
AI Technical Summary
Traditional ultrasound diagnostic techniques suffer from imbalances in medical resources, insufficient diagnostic efficiency, and unstable imaging quality. Furthermore, they heavily rely on physicians manually measuring lesion parameters, resulting in high subjectivity and low efficiency.
A real-time segmentation method for cardiac ultrasound based on Unet networks is adopted, which combines encoder-decoder symmetrical network topology, skip connection mechanism, transposed convolutional layer and hollow spatial pyramid pooling module to achieve real-time segmentation and automatic measurement of lesion area, and supports access to mainstream ultrasound equipment and intelligent auxiliary diagnosis of software system.
It has significantly improved the efficiency and accuracy of cardiac ultrasound diagnosis, reduced the workload of doctors, optimized the allocation of medical resources, ensured high-quality ultrasound examinations in remote areas, and promoted the development of smart healthcare.
Smart Images

Figure CN122454575A_ABST