Back to results
US 2021/0233301 A1Published application
Orientation detection in fluoroscopic images
AssigneeCovidien LP
InventorsDaniel Ovadia, Guy Alexandroni, Ariel Birenbaum
Life of this patent
Estimating the term…
Abstract
Systems and methods of automatic orientation detection in fluoroscopic images using deep learning enable local registration for correction of initial CT-to-body registration in Electromagnetic Navigation Bronchoscopy (ENB) systems.
Claims20
A method of performing local registration in a medical device navigation procedure, the method comprising:
- acquiring fluoroscopic images of at least one anatomic feature and a medical device in a patient's body captured by a fluoroscopic imaging device;
- processing the fluoroscopic images by a neural network;
- to
- determining the orientation of the patient's body in the fluoroscopic images based on the processing of the fluoroscopic images by the neural network;
- and
- generating a three-dimensional (3D) reconstruction of the fluoroscopic images based on at least the determined orientation of the patient's body in the fluoroscopic images.
3 more elements
A system for performing local registration using fluoroscopy comprising:
- a processor in communication with a fluoroscope;
- and
- a memory configured to store a neural network and instructions that, when executed by the processor, cause the processor to:
- acquire fluoroscopic images of at least an anatomical structure and a medical device in a patient's body;
- process the fluoroscopic images with a neural network;
- determine the orientation of the patient's body in the fluoroscopic images based on the results of processing the fluoroscopic images with the neural network;
- generate a three-dimensional (3D) reconstruction of the fluoroscopic images based on at least the orientation of the determined orientation candidate;
- and perform local registration based on the 3D reconstruction.
5 more elements
A method for detecting an orientation of a patient's body in a sequence of frames of fluoroscopic images, the method comprising:
- acquiring the sequence of frames of the fluoroscopic images from a fluoroscopic sweep of the patient's body;
- processing the fluoroscopic image frames with a convolutional neural network to obtain probability values for orientation candidates;
- for each fluoroscopic image frame in a subset of the frames of fluoroscopic images, selecting an orientation candidate having a highest probability value;
- and
- determining that the orientation of the patient's body corresponds to the orientation candidate that is selected the greatest number of times.
2 more elements
Dependencies are inferred from the claim text.