patent.fyi
Applications  /  US 2026/0354781
United States Patent Application Publication

US 2026/0354781 A1

Published
Nov 12, 2026
Filed
May 8, 2026
Earliest date on the face
May 9, 2025
Classification
A61B 6/487
First published here
Nov 19, 2026

Generated automatically. Not reviewed by a person. Not legal advice or an opinion.

Section 1 · machine-generated

What if?

Questions about how individual parts of claim 1 might be changed. A question may describe something different from what the claim language recites, or a version of it; no view is offered on which.

1.1Stationary exposures at indexed positions

What ifWhat if the imaging assembly stops at discrete positions and captures every frame while stationary?
Claim language“one or more continuous translational sweeps”

A motorized carriage moves the source and detector between indexed positions and settles before each exposure. At each anatomical station, the assembly also stops at several angular positions to acquire views for depth recovery, and neighboring stations share a band of anatomical coverage.

1.2Encoder based imaging geometry

What ifWhat if carriage and gantry encoders determine imaging pose without a mounted imaging marker?
Claim language“a first fiducial marker rigidly mounted”

Linear and rotary encoders report the carriage position and gantry angles at each exposure. A calibrated kinematic model converts these readings into source and detector poses in the table coordinate system, and a table-referenced tracker supplies the patient pose.

1.3Markerless patient surface tracking

What ifWhat if a depth camera tracks the patient surface without an attached patient marker?
Claim language“a second fiducial marker affixable to the patient”

A calibrated depth camera captures an exposed surface patch over a supported anatomical segment and uses the first point cloud as the patient reference. At each exposure, rigid point-cloud registration estimates the motion of the patch, and the processor combines that transform with the imaging pose to express the projection in patient coordinates.

1.6Forward projection with simplex search

What ifWhat if the processor fits an anatomical model using only forward-rendered radiographs and samples the fitted model onto a voxel grid?
Claim language“back-projects the one or more two-dimensional representations”

The processor uses the frame projection matrices to render radiographs of a statistical attenuation model and scores each rendering against its measured frame. A simplex search updates a finite set of shape and pose coefficients using the aggregate similarity score. The processor samples the fitted attenuation model at voxel centers in patient coordinates.

1.7Regularized iterative volume reconstruction

What ifWhat if a regularized iterative solver produces the volume from the back-projected data?
Claim language“a trained three-dimensional deep learning network”

The processor initializes the volume with normalized back-projected radiographic intensities. Each iteration forward-projects the current estimate, back-projects the measured-minus-predicted residuals, and applies a nonnegativity constraint and a three-dimensional total variation penalty.

Section 2 · machine-generated

Variations

Further arrangements of the same subject matter, described here as of the publication date at the foot of this page.

Other field of use

Volumetric inspection of composite spar bonds

A motorized carriage sweeps an X-ray source and flat-panel detector along a composite spar at two oblique orientations, collecting overlapping radiographs at each orientation. An optical tracker measures one fiducial rigidly attached to the imaging head and another fixed to the spar, and calibrated projection matrices express each view in spar coordinates. The processor back-projects the radiographs into a voxel grid and applies a three-dimensional network trained on scan volumes and sectioned calibration coupons to estimate the distribution of bond-line voids.

Component substitution

Electromagnetic tracking of fluoroscopic poses

A low-frequency electromagnetic field generator replaces the optical tracking cameras, and two six-degree-of-freedom coil sensors serve as fiducials on the imaging head and near the anatomy. A calibration fixture establishes the rigid transform between the head sensor and the fluoroscopic projection geometry, and a common clock timestamps tracking samples and image exposures. The processor interpolates both sensor poses at each exposure, composes their relative transform with the calibrated projection geometry, and uses the resulting matrices for back-projection before three-dimensional inference.

Parameter or scale variation

Two passes with fixed pulse spacing

A motorized carriage travels 420 mm at 30 mm/s on each of two passes with the fluoroscopic head set at minus 20 degrees and plus 20 degrees from the anteroposterior orientation. The head acquires 71 frames per pass at 5 pulses per second, including a pulse at each endpoint, giving 6 mm between successive exposure centers along the sweep. A tracking system measures the head and patient fiducial poses at each exposure, and the two sets of calibrated projections populate a common 1 mm isotropic patient-coordinate voxel grid for three-dimensional inference.

Combination with existing systems

Reconstructed anatomy in surgical navigation

A surgical navigation workstation reads the patient reference fiducial used for the fluoroscopic sweep and tracks a separate fiducial rigidly mounted to a surgical probe. The reconstruction processor sends the volume and its voxel-to-patient transform to the workstation. A pivot calibration locates the probe tip relative to its fiducial, and the workstation combines live tracking with the inverse voxel-to-patient transform to display the tip and shaft on corresponding slices.

Automated or data-driven variant

Automatic rescanning of underobserved regions

After an initial sweep, acquisition software projects each voxel center in the prescribed reconstruction region into every calibrated frame and records the viewing directions for centers that fall within the exposed detector area. The software flags contiguous regions for which no pair of recorded rays differs by at least 15 degrees and scores available oblique carriage trajectories by how many flagged voxels they cover at the required angular separation. The carriage runs the highest-scoring trajectory, and the processor combines the new and original projections in the same patient-coordinate grid before running the three-dimensional network.

Automatic masking of transient image obstructions

A two-dimensional segmentation network trained on annotated fluoroscopic frames produces a per-pixel mask of transient objects such as handheld tools and retractors. During back-projection, the processor gives masked detector samples zero weight and accumulates both the weighted image values and the contributing observation count at each voxel. A three-dimensional network receives the normalized intensity grid and the observation-count grid as separate channels and generates the reconstructed volume in patient coordinates.

Specific version of a part

Manual registration with paired landmarks

The operator identifies at least four well-separated radiopaque landmarks in the reconstructed volume and digitizes their matching physical locations with a tracked pointer referenced to the patient. The processor calculates a least-squares rigid transform from the paired points and applies it to the volume.

Pulsed sweep with encoder triggering

A motorized carriage translates the source and detector continuously along the table while a linear encoder triggers one exposure every 20 millimeters. The detector captures more than four frames, and a 160-millimeter longitudinal footprint at the reference plane provides 140 millimeters of overlap there between consecutive frames.

Bolted optical marker array

A machined aluminum bracket carries four retroreflective spheres in an asymmetric, noncoplanar pattern. Dowel pins locate the bracket on the detector housing, bolts hold it in place, and calibration relates the marker coordinate system to the detector plane and source position.

Synchronized stereo pose tracking

Two calibrated infrared cameras triangulate the sphere centers on each marker array and fit each measured set to its known rigid geometry to recover translation and rotation. Exposure timestamps align the two poses with each fluoroscopic frame, and the relative pose, imaging calibration, and detector intrinsics determine the projection matrix.

Per frame multiscale feature maps

The processor applies flat-field correction and intensity normalization to each fluoroscopic frame before passing it to a trained two-dimensional convolutional encoder. The encoder returns feature maps at one-quarter and one-eighth of the input resolution, and the processor records the pixel-to-feature coordinate transform for each scale.

Section 3 · quoted from USPTO records

Other work

Passages reproduced word for word from two earlier-published U.S. patent documents, placed beside the language of claim 1. No characterization is offered. Each passage is marked A or B for the document it comes from; ¶ numbers are the paragraph numbers in the USPTO’s published full text of that document.

Document AUS 7,103,136 B2
Published
Sep 5, 2006
Filed
Dec 22, 2003
Class
A61B 6/03
Document BUS 11,399,895 B2
Published
Aug 2, 2022
Filed
Feb 10, 2017
Class
A61B 34/20
Claim 1
Document A
Document B
1.1(a) a fluoroscopic image acquisition system configured to capture a sequence of N fluoroscopic frames during one or more continuous translational sweeps along an extended anatomical region of interest, wherein N>4, and wherein successive frames have at least partial spatial overlap;
A method for generating a three-dimensional image, comprising: acquiring two or more fluoroscopic projection images at different view angles relative to a region of interest, wherein the two or more fluoroscopic projection images are aligned using at least one of a fiducial marker or an anatomical structure; and reconstructing at least one three-dimensional image from the two or more fluoroscopic projection images.
A · US 7,103,136 B2 · claim 1
The images may be obtained at discrete locations, but, optionally, frames from a continuous fluoroscopic video can be used for reconstruction.
B · US 11,399,895 B2 · ¶ 96
1.2(b) a first fiducial marker rigidly mounted to the fluoroscopic image acquisition system;
Intrinsic calibration, which is the process of correcting image distortion in a received image and learning the projective geometry of the imager, involves placing fiducial markers 400 (also known as calibration markers) in the path of the X-ray, where a fiducial marker is an object opaque to X-rays. The fiducial markers 400 are rigidly arranged in predetermined patterns in one or more planes in the path of the X-rays and are visible in the recorded images.
B · US 11,399,895 B2 · ¶ 74
1.3(c) a second fiducial marker affixable to the patient proximate the anatomical region of interest;
An optical tracking system uses a position sensor to detect infrared-emitting or retro-reflective markers attached to the teleoperational assembly, the medical instrument, and/or patient.
B · US 11,399,895 B2 · ¶ 68
1.4(d) a spatial tracking system comprising at least one spatial sensing device and configured to track both the first and second fiducial markers during at least a portion of the sequence of fluoroscopic frames, thereby generating six-degree-of-freedom positional data for at least one frame in the sequence, the positional data being usable, in combination with intrinsic camera parameters of the fluoroscopic image acquisition system, to derive a projection matrix, wherein a projection matrix for each of the remaining frames is derived from the positional data by at least one of: direct tracking measurement, interpolation between tracked frames, image-based inter-frame registration, or any combination thereof;
The projection images can be registered (or the acquisition geometry for each frame determined) from fiducial marks if available or anatomical markers present in the image.
A · US 7,103,136 B2 · ¶ 43
1.5(e) a processor configured to execute a trained neural network model that: (i) produces, from the N fluoroscopic frames, one or more two-dimensional representations, each two-dimensional representation being selected from: a raw fluoroscopic image, a composited panoramic image formed by spatially registering and blending a plurality of sequential frames, or learned feature maps extracted by a trained two-dimensional deep learning network operating on individual fluoroscopic frames or on a composited panoramic image;
Raw images generated by X-ray detector 386 tend to suffer from undesirable distortion (e.g., “barrel distortion” and/or “S-distortion”) caused by a number of factors, including inherent image distortion in the image intensifier and external electromagnetic fields.
B · US 11,399,895 B2 · ¶ 74
1.6(e) a processor configured to execute a trained neural network model that: (ii) back-projects the one or more two-dimensional representations into one or more three-dimensional voxel grids using projection matrices derived from the positional data; and
Reconstruction of the acquired fluoroscopic image data into a three-dimensional image or volume may be accomplished by various processes. For example, the reconstruction may be accomplished by backprojection, filtered backprojection, algebraic reconstruction technique (ART), or any other reconstruction method known in the art.
A · US 7,103,136 B2 · ¶ 43
1.7(e) a processor configured to execute a trained neural network model that: (iii) produces a volumetric output by processing the back-projected representations through a trained three-dimensional deep learning network; and
Three-dimensional tomographic images can be generated from the two-dimensional projection images created from different projection angles or viewpoints.
B · US 11,399,895 B2 · ¶ 75
1.8(e) a processor configured to execute a trained neural network model that: (f) wherein the volumetric output spans a spatial extent, in at least one dimension, exceeding the field of view of any single fluoroscopic frame, and is automatically registered to a patient coordinate system via spatial tracking measurements of the first and second fiducial markers, the registration being independent of any preoperative volumetric imaging study of the patient and any manual intraoperative landmark-correspondence step.

An empty cell means only that no passage was reproduced there.

Statements in these documents
In addition, the anatomical context of the fluoroscopic images may be difficult to ascertain without a three-dimensional references.
A · US 7,103,136 B2 · ¶ 3
The projection images acquired at the various perspectives may be utilized to reconstruct a three-dimensional image of the region which, when displayed, provides a three-dimensional anatomic context for the region of interest.
A · US 7,103,136 B2 · ¶ 4
The present disclosure is directed to systems and methods for conducting an image guided procedure, and more particularly to systems and methods for pose estimation, calibration of a perspective imaging system, and real-time tomosynthesis to enhance the accuracy of tool navigation during an image guided procedure.
B · US 11,399,895 B2 · ¶ 2
With the image-guided instruments registered to the images, the instruments may navigate natural or surgically created passageways in anatomical systems such as the lungs, the colon, the intestines, the kidneys, the heart, the circulatory system, or the like.
B · US 11,399,895 B2 · ¶ 3
Cite aspatent.fyi, US 2026/0354781 A1, claim 1, first published 2026-11-19T03:31:31-05:00, https://patent.fyi/US20260354781/, SHA-256 bd96ce891d428b0c177862b11280c0af7d1b0cb356e2489708c7ab3619c3fe68
First published
Permanent addresshttps://patent.fyi/US20260354781/
Content fingerprint (SHA-256)bd96ce891d428b0c177862b11280c0af7d1b0cb356e2489708c7ab3619c3fe68
Methodv2026-11-18.2 · automated, not reviewed by a person
Correctionscorrections@patent.fyi