US-12616440-B2 - Robust view classification and measurement in ultrasound imaging

US12616440B2US 12616440 B2US12616440 B2US 12616440B2US-12616440-B2

Abstract

For robust view classification and measurement estimation in sequential ultrasound imaging, the classification and/or measurements for a given image or sequence of images are gated. To prevent oscillation in results, the gating provides consistent output.

Inventors

Dominik Neumann
Awais MANSOOR
Vlad Comanelea-Serban
Sasa Grbic
Mallory Selzo
Andrzej Milkowski

Assignees

SIEMENS MEDICAL SOLUTIONS USA, INC.

Dates

Publication Date: 20260505
Application Date: 20241205

Claims (20)

1 . A method for sequential imaging with an ultrasound scanner, the method comprising: scanning a patient with the ultrasound scanner, the scanning resulting in a sequence of frames of ultrasound data representing the patient; classifying a view for at least some of the frames of ultrasound data of the sequence; gating the classified views from multiple of the frames of ultrasound data, the gating of the classified views including: rejecting a first one of the classified views where a first classification from the first one of the classified views implies a jump in location by the ultrasound scanner; and passing a second one of the classified views where a second classification from the second one of the classified views implies a continuous transition by the ultrasound scanner; automatically measuring from the frames of ultrasound data, including the second one of the classified views, that pass the gating of the classified views; determining confidence levels for results of the measuring; gating the results of the measuring based on the confidence levels; and displaying images of the frames of data, the images including annotation of the classified views passed by the gating of the classified views, the images including the results of the measuring passed by the gating of the results of the measuring, the displaying being free of annotations from any of the frames of ultrasound data rejected by the gating of the classified views and being free of results of the measuring rejected by the gating of the results of the measuring.
2 . The method of claim 1 wherein scanning comprises scanning the patient in an abdominal examination, wherein measuring comprises measuring a distance, area, or volume.
3 . The method of claim 1 wherein gating the classified views further comprises passing the classified view where a threshold number or percentage of the classified views in a window of classified views are a same classified view and rejecting the different classified views in the window.
4 . The method of claim 3 wherein gating the classified views further comprises determining a confidence for each of the classified views, and only passing the classified where the threshold number or percentage of the classified views in the window are the same classified view and have a confidence above a confidence threshold.
5 . The method of claim 1 wherein classifying the view further comprises concatenating the multiple frames of ultrasound data in different windows and classifying the concatenations of the multiple frames from the different windows, and wherein gating the classified views comprises gating the classified views from the concatenations.
6 . The method of claim 1 wherein gating the classified views further comprises gating as a dynamic system with state change.
7 . The method of claim 1 wherein gating the classified views further comprises storing the classified views or the frames of ultrasound data in a first-in-first-out queue and gating based on the classified views or the frames in the first-in-first-out queue.
8 . The method of claim 1 wherein gating the results comprises gating based on quality of the frames of ultrasound data, respectively.
9 . The method of claim 1 wherein gating the results comprises gating based on an amount of motion.
10 . The method of claim 1 wherein gating the results comprises gating based on measurement stability of the results.
11 . The method of claim 10 wherein gating the results comprises aggregating the results when the measurement stability is above a stability threshold.
12 . The method of claim 1 wherein measuring comprises measuring from a concatenation of the frames that pass the gating of the classified views, and wherein gating the results comprises gating based on the concatenation.
13 . The method of claim 1 wherein gating the results comprises spatial-temporal tracking.
14 . A method for view classification and measurement in ultrasound imaging by an ultrasound scanner, the method comprising: imaging, by the ultrasound scanner, a patient as part of an abdominal examination, the imaging generating images of an abdomen of the patient; classifying, by the ultrasound scanner, structure represented in the images; measuring, by the ultrasound scanner, the structure; and displaying only the classification consistent with smooth translation of the ultrasound scanner and displaying only consistent measurement of the structure.
15 . The method of claim 14 , wherein displaying comprises gating the classifying and gating the measuring such that display of any deviating structure and any deviating measurement is suppressed.
16 . The method of claim 15 wherein gating comprises gating the classification further with temporal filtering, sequence-level classification, or dynamic system modeling and gating the measurement with image quality, motion detection, measurement stability, measurement aggregation, sequence-level measurement, or spatial-temporal tracking.
17 . A system for classification and measurement in ultrasound imaging, the system comprising: an ultrasound imager configured to scan with ultrasound; a processor configured to classify views from the scan, measure structure in the views, and determine which of the classifications and measurements to display, the determination of which classifications to display is based on whether the classifications are consistent with progression of the scan without jumping; and a display configured to: display, based on instructions from the processor, first classifications that are consistent with progression of the scan without jumping and first measurements determined by the processor to display; and not display, based on instructions from the processor, second classifications that indicate a jump within the scan and second measurements determined by the processor to not display.
18 . The system of claim 17 , wherein the processor is configured to determine which of the measurements to display based on a level of consistency through a sequence of the scan.
19 . The system of claim 17 , further comprising a memory configured to store the scan, and wherein the processor is configured to classify, measure, and determine from the stored scan.
20 . The system of claim 17 , wherein the processor is further configured to determine which of the classifications and measurements to display by further performing spatial-temporal tracking.

Description

PRIORITY This application claims priority to and is a continuation of U.S. patent application Ser. No. 17/325,257, filed May 20, 2021, which is incorporated by reference herein in its entirety. BACKGROUND The present embodiments relate to ultrasound imaging. In abdominal ultrasound examinations, several minutes of real-time imaging at about 20 frames per second is performed, which results in thousands of individual frames to cover all relevant views. The sonographer typically follows a medical guideline to examine organs or other anatomical structures of interest, thereby generating a continuous stream of sequential imaging data. During the scan, the sonographer manually selects “key images,” which are then stored with an identified view label (e.g., “left kidney transversal”, “common bile duct”, . . . ). For some of these key images, the sonographer further measures certain quantities of interest (e.g., kidney length, width, and/or height, lumen diameter, gallbladder wall thickness, . . . ). Measurement is typically done after the user found a proper view (e.g., high image quality, organ coverage, . . . ), then “freezes” the image, and finally places measurement calipers on the image. Key image selection, view identification, image “freezing”, and measurement caliper placement are done manually on-the-fly while scanning. These manual steps are prone to user error and may be inconsistent across sonographers (e.g., typos, different abbreviations, inter-user variability). The quality of the results depends on user experience and skill. Given the number of images, the overall process can be inefficient and tedious. Image processing, such as based on artificial intelligence, attempts to automate some of these manual tasks. An individual image is input to perform one or more of these tasks on that image. This approach may lead to unstable predictions image-by-image. The user typically selects the key image to input, and there is no guarantee that the one frame the user selected gives the optimal result by the artificial intelligence. SUMMARY By way of introduction, the preferred embodiments described below include methods, computer readable storage media, instructions, and systems for robust view classification and measurement estimation in sequential ultrasound imaging. The classification and/or measurements for a given image or sequence of images are filtered and/or gated. To prevent oscillation in results, the filtering and/or gating provides consistent output. In a first aspect, a method is provided for sequential imaging with an ultrasound scanner. A patient is scanned with the ultrasound scanner. The scanning results in a sequence of frames of ultrasound data representing the patient. A view is classified for at least some of the frames of ultrasound data of the sequence. The classified views from multiple of the frames of ultrasound data are gated, where the gating of the classified views rejects a first one of the classified views and passes a second one of the classified views. Structure in the frames of ultrasound data are automatically measured. The frames where measurement is performed are the frames that pass the gating of the classified views. Results of the measuring from the frames that pass the gating of the classified views are gated. The gating of the results of the measuring rejects a first one of the results of the measuring and passes a second one of the results of the measuring. Images of the frames of data are displayed. The images include annotation for the classified views passed by the gating of the classified views. The images include the results of the measuring passed by the gating of the results of the measuring. The display is free of classified views from any of the frames of ultrasound data rejected by the gating of the classified views and free of the results of the measures rejected by the gating of the results of the measuring. In a second aspect, a method is provided for view classification and measurement in ultrasound imaging by an ultrasound scanner. The ultrasound scanner images a patient as part of an abdominal examination. The imaging generates images of an abdomen of the patient. Structure represented in the images is classified. The structure is measured. Only the images with consistent classification of the structure and consistent measurement of the structure are displayed. In a third aspect, a system is provided for stable classification and measurement in ultrasound imaging. An ultrasound imager is configured to scan with ultrasound. A processor is configured to classify views from the scan, measure structure in the views, and determine which of the classifications and measurements to display. A display is configured to display first images for the classifications and measurements determined by the processor to display and not display the classification and measurements determined by the processor to not display. The present invention is defined by the following