Method and/or system for magnetic localization

Patent metadata

Field	Value
Publication number	US-11250318-B2
Application number	US-201414889578-A
Country	US
Kind code	B2
Filing date	May 7, 2014
Priority date	May 7, 2013
Publication date	Feb 15, 2022
Grant date	Feb 15, 2022

Abstract

A method of real time magnetic localization comprising: providing an artificial neural network field model that is calibrated and optimized for a predetermined magnet; receiving signals from one or more magnetic sensors; and solving the location of the magnet using the model based on the signals.

First claim

The invention claimed is:

1. A method of real time in vivo location determination comprising:
providing a 2-input-2-ouput artificial neural network (ANN) field model for predicting magnetic vector fields of a predetermined magnet with an axisymmetric magnetic vector field, each predicted magnetic vector field having two cylindrical prediction coordinates;
inserting a medical instrument into a patient, the medical instrument being associated with a magnetic sensor, and the magnetic sensor being associated with the predetermined magnet;
receiving, from the magnetic sensor, a sensor signal representing a measured axisymmetric magnetic vector field of the predetermined magnet, the measured axisymmetric magnetic vector field having three Cartesian measurement coordinates;
converting the three Cartesian measurement coordinates of the measured axisymmetric magnetic vector field into two cylindrical measurement coordinates;
providing the two cylindrical measurement coordinates as the two inputs to the 2-input-2-output ANN field model;
estimating- a relative position between the predetermined magnet and the magnetic sensor using the 2-input-2-output ANN field model by minimizing a cost function to produce two cylindrical coordinates as the two outputs of the 2-input-2-output ANN field model, the cost function comprising a difference between the two cylindrical measurement coordinates of the measured axisymmetric magnetic vector field and the two cylindrical prediction coordinates of each predicted magnetic vector field predicted by the 2-input-2-output ANN field model; and

converting the two cylindrical coordinates to corresponding three Cartesian coordinates for identifying a location of the medical instrument within the patient.

2. The method of claim 1 wherein the location of the medical instrument is determined when the magnetic sensor is adjacent the magnet.

3. The method of claim 2 wherein the determined location has an accuracy of less than 1 mT (RMSE) and/or a peak absolute error of less than 40 mT when the magnet is located within 25 mm from the magnetic sensor.

4. The method of claim 1 wherein the medical instrument includes the magnet, the method further comprising:
arranging the magnetic sensor on or about the patient.

5. The method of claim 1 wherein the model is based on a back propagation neural network field model trained using a Levenberg-Marquardt supervised learning algorithm.

6. The method of claim 1 further comprising selecting an order of the model and/or a number of nodes to reduce the error below a predetermined threshold.

7. The method of claim 6 wherein the order is a single hidden layer and the number of nodes is 5 to 20 hidden nodes.

8. The method of claim 1 further comprising providing a plurality of weighting coefficients for the model, wherein the weighting coefficients are pre-optimized for the predetermined magnet.

9. The method of claim 1 wherein the magnet is passive and untethered.

10. The method of claim 1 , wherein the medical instrument includes the magnetic sensor, the method further comprising:
arranging the magnet on or about the patient.

11. The method of claim 1 , wherein providing the ANN field model comprises:
obtaining a magnetic vector field map of the magnet;
extracting an axis-symmetric field slice from the obtained magnetic vector field map to train or fit the ANN field model; and
determining a number of associated hidden nodes in accordance with a residue error based on the extracted axis-symmetric field slice.

12. The method of claim 1 , wherein the cost function, ‘C’, is:
C=Σ∥B model ( x S )− B measured ∥ 2

where ‘B model (X S )’ and ‘B measured ” represent a model predicted magnetic vector field where the magnetic vector field where the magnetic sensor is located at ‘x S ’, respectively.

13. The method of claim 1 , wherein the medical instrument is further associated with a further magnetic sensor, the further magnetic sensor is associated with the predetermined magnet, and the method further comprises:
receiving, from the further magnetic sensor, a further sensor signal representing a further measured axisymmetric magnetic vector field of the magnet, the further measured axisymmetric magnetic vector field having three Cartesian measurement coordinates; and
converting the three Cartesian measurement coordinates of the further measured axisymmetric magnetic vector field into two cylindrical measurement coordinates,
wherein the cost function further comprises a difference between the two cylindrical measurement coordinates of the further measured axisymmetric magnetic vector field and the two cylindrical prediction coordinates of each predicted magnetic vector field.

14. A system for real time in vivo location determination, comprising:
a magnet with an axisymmetric magnetic vector field;
a magnetic sensor configured to be associated with the magnet and to measure the axisymmetric magnetic vector field of the magnet to provide a sensor signal;
a medical instrument configured to be inserted into a patient and to be associated with the magnetic sensor;
a processor configured to be associated with the magnetic sensor;
a storage device configured to be associated with the processor and to store software instructions for causing the processor to:
provide 2-input-2-output artificial neural network (ANN) field model for predicting magnetic vector fields of the magnet, each predicted magnetic vector field having two cylindrical prediction coordinates;
receive, from the magnetic sensor, the sensor signal representing the measured axisymmetric magnetic vector field of the magnet, the measured axisymmetric magnetic vector field having three Cartesian measurement coordinates;
convert the three Cartesian measurement coordinates of the measured axisymmetric magnetic vector field into two cylindrical measurement coordinates;
provide the two cylindrical measurement coordinates as the two inputs to the 2-input-2-output ANN field model;

estimate a relative position between the magnet and the magnetic sensor using the 2-input-2-output ANN field model by minimizing a cost function to produce two cylindrical coordinates as the two outputs of the 2-input-2-output ANN field model, the cost function comprising a difference between the two cylindrical measurement coordinates of the measured axisymmetric magnetic vector field and the two cylindrical prediction coordinates of each predicted magnetic vector field predicted by the 2-input-2-output ANN field model;
convert the two cylindrical coordinates to corresponding three Cartesian coordinates for identifying a location of the medical instrument; and
a display or indicator to show the identified location of the medical instrument within the patient.

15. The system of claim 14 wherein the magnet is passive and untethered.

16. The system of claim 15 wherein the magnet is an axially magnetized Neodymium annular cylinder permanent magnet.

17. The system of claim 14 wherein the model includes weighting coefficients, an order and/or the number of nodes for the magnet and/or a desired error threshold.

18. The system of claim 14 wherein the medical instrument includes the magnet, and the magnetic sensor is further configured to be arranged on or about the patient.

19. The system of claim 14 wherein the medical instrument includes the magnetic sensor, and the magnet is configured to be arranged on or about the patient.

20. The system of claim 14 , wherein the software instructions corres

Assignees

• Univ Singapore Technology & Design
• Massachusetts Inst Technology
• Massachusetts Institute Of Tech Mit

Inventors

• Foong Shaohui
• Wu Faye

Classifications

• G06N3/09 Primary

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G01B7/003 Primary

  for measuring position, not involving coordinate determination (coordinate measuring G01B7/004) · CPC title

• G06N3/063

  using electronic means · CPC title

• A61B5/7264

  Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems · CPC title