Machine learning based real-time radiation dose assessment

Patent metadata

Field	Value
Publication number	US-10645259-B2
Application number	US-201715688960-A
Country	US
Kind code	B2
Filing date	Aug 29, 2017
Priority date	Aug 31, 2016
Publication date	May 5, 2020
Grant date	May 5, 2020

Abstract

A computer-implemented method for performing radiation dose assessment includes acquiring a plurality of parameters related to an image scanner configured to apply a heat or radiation dose to a subject during imaging. A plurality of features is extracted from the plurality of parameters. A dose map of a region of interest in the room is generated by applying a machine learning model to the plurality of features and the interventional room scene.

First claim

The invention claimed is:

1. A computer-implemented method for performing radiation dose assessment, the method comprising:
acquiring a plurality of parameters related to an image scanner configured to apply a heat or radiation dose to a subject during imaging in a room;
extracting a plurality of features from the plurality of parameters;
generating a dose map of a region of interest in the room by applying a machine learning model to the plurality of features wherein the dose map comprises a visualization of the room overlaid by a heat map indicating areas of high dosage;
presenting the dose map on a display;
monitoring a location of an operator within the room; and
in response to detecting that the operator is within an area of high dosage, generating an alarm for the operator.

2. The method of claim 1 , wherein the region of interest is specified by a user of the image scanner.

3. The method of claim 1 , further comprising:
monitoring a location of an operator within the room;
determining a time period that the operator has been within one or more of the areas of high dosage;
if the time period exceeds a predetermined threshold, generating an alarm for the operator.

4. The method of claim 1 , further comprising:
creating a projection of the dose map on a floor area of the room, wherein the projection indicates high dose area and low does areas on the floor area.

5. The method of claim 1 , further comprising:
detecting a change to the plurality of parameters related to the image scanner;
in response to detecting the change, extracting a plurality of updated features from the plurality of parameters; and
generating an updated dose map of the room by applying the machine learning model to the plurality of updated features.

6. The method of claim 1 , further comprising:
determining an interventional room scene corresponding to a room in which the image scanner is operating, wherein
the dose map of the region of interest in the room is generated by applying a machine learning model to the plurality of features and the interventional room scene.

7. The method of claim 6 , further comprising:
determining an update to the interventional room scene corresponding to the room in which the image scanner is operating; and
generating an updated dose map of the room by applying the machine learning model to the plurality of features and the updated interventional room scene.

8. The method of claim 6 , wherein the interventional room scene is determined by imaging the room with an RGB camera and a depth sensor.

9. The method of claim 1 , further comprising:
receiving a measured dose experienced by the operator via a dosimeter within the room;
determining a predicted dose for the operator based on the dose map;
determining an error between the measured dose and the predicted dose; and
updating the plurality of parameters based on the error.

10. A computer-implemented method for performing radiation dose assessment, the method comprising:
training a machine learning model to map features that characterize a geometric model of a dose metric, wherein the features comprises geometric measurements of a subject's body surface and internal organs;
during an imaging procedure, using the machine learning model to generate a dose map of a room in which the imaging procedure is being performed, wherein input to the machine learning model comprises (a) features related to an image scanner configured to apply a heat or radiation dose to a subject during the imaging procedure and (b) an interventional room scene corresponding to a room in which the image scanner is operating.

11. The method of claim 10 , wherein the geometric model further comprises an image scanner, geometric model and the features comprise one or more of tube current, tube voltage, collimation, C-arm angle, exposure, distance to an imaging table, and distance to iso-center of the image scanner.

12. The method of claim 10 , wherein the geometric model further comprises an anatomical model of operators of the image scanner and the features comprise one or more of a number of operators, operator locations in the room, and an indication of shielding device worn by the operators.

13. The method of claim 10 , wherein the geometric model further comprises geometric models of non-scanner equipment located in the room and the features comprise geometric features characterizing the shape, size, and configuration of the room.

14. The method of claim 10 , further comprising:
performing a plurality of Monte Carlo simulations for radiation transport at a plurality of operating points of the image scanner to generate ground truth data, wherein the plurality of operating points comprise one or more of tube current, voltage, collimation, and distance to iso-center of the image scanner,
wherein the machine learning model is trained using the ground truth data.

15. The method of claim 14 , wherein the plurality of Monte Carlo simulations are performed for a plurality of room configurations.

16. The method of claim 14 , wherein the plurality of Monte Carlo simulations are performed for a plurality of patient models.

17. The method of claim 14 , further comprising:
acquiring a plurality of images of a phantom using the image scanner;
during acquisition of the images, acquiring a plurality of dose measurements using a plurality of dose sensors distributed in the room,
wherein the machine learning model is trained using the plurality of dose measurements.

18. The method of claim 14 , further comprising:
generating a plurality of 3D mesh models representative of a plurality of anatomical organs based on 3D image data acquired from a plurality of subjects;
for each 3D mesh model, using a physics based radiation dose computation engine to compute a plurality of ground truth dose metrics for a plurality of parameters associated with the image scanner,
wherein the machine learning model is trained using the ground truth dose metrics.

19. The method of claim 14 , further comprising:
generating a plurality of synthetic anatomical models representative of a plurality of subjects;
for each synthetic anatomical model, using a physics based radiation dose computation engine to compute a plurality of ground truth dose metrics for a plurality of parameters associated with the image scanner,

wherein the machine learning model is trained using the ground truth dose metrics.

Assignees

• Siemens Healthcare Gmbh

Inventors

• Sharma Puneet
• Rapaka Saikiran

Classifications

• G01T1/02 Primary

  Dosimeters (G01T1/15 takes precedence) · CPC title

• H04N5/3572

  Electricity · mapped topic

• H04N5/232

  Electricity · mapped topic

• H04N5/217 Primary

  Electricity · mapped topic

• H04N23/60

  Control of cameras or camera modules · CPC title