US20260224304
2026-08-06
Human necessities
A61B34/20
The patent application describes a system and method for in vivo navigation of medical devices using machine-learning. The approach involves receiving medical imaging data of a patient's anatomy, along with non-optical in vivo image data from a sensor on the device. A trained model, developed using medical and non-optical imaging data from multiple individuals, is used to locate the device's distal end in the imaging data. This allows for the modification of the input imaging data to depict the device's location, which is then displayed to the user.
This innovation pertains to machine-learning techniques for in vivo navigation, particularly focusing on the integration of non-optical image data, such as ultrasound, with medical imaging data. The goal is to enhance the precision of medical procedures by accurately determining the position of devices within a patient's body, thereby addressing challenges associated with traditional invasive techniques.
The system comprises a memory, display, and processor. The memory stores instructions and a trained machine-learning model, which associates non-optical in vivo image data with medical imaging data. The processor executes these instructions to receive and process input data, determine the device's location, and modify the imaging data to include a location indicator. This updated data is then displayed, providing real-time navigation assistance.
The system can provide real-time updates as the medical device moves within the anatomy. This involves receiving additional non-optical image data, determining an updated location, and adjusting the location indicator accordingly. The display outputs these updates, allowing for continuous monitoring of the device's position during medical procedures.
The machine-learning model is capable of learning associations between sequences of images and paths of travel, enhancing its predictive capabilities. It can also incorporate additional data, such as position signals and three-dimensional structures, to refine location determination. The model supports various data types, including ultrasound and 360-degree imaging, and can adapt to different anatomical regions, such as the peripheral lung areas.