System and method for facilitating DBS electrode trajectory planning

Patent metadata

Field	Value
Publication number	US-11786309-B2
Application number	US-202017135022-A
Country	US
Kind code	B2
Filing date	Dec 28, 2020
Priority date	Dec 28, 2020
Publication date	Oct 17, 2023
Grant date	Oct 17, 2023

Abstract

A system and method for facilitating DBS electrode trajectory planning using a machine learning (ML)-based feature identification scheme configured to identify and distinguish between various regions of interest (ROIs) and regions of avoidance (ROAs) in a patient's brain scan image. In one arrangement, standard orientation image slices as well as re-sliced images in non-standard orientations are provided in a labeled input dataset for training a CNN/ANN for distinguishing between ROIs and ROAs. Upon identification of the ROIs and ROAs in the patient's brain scan image, an optimal trajectory for implanting a DBS lead may be determined relative to a particular ROI while avoiding any ROAs.

First claim

The invention claimed is:

1. A computer-implemented method of electrode trajectory planning for deep brain stimulation (DBS), the method comprising:
obtaining a set of medical imaging data pertaining to human cranial anatomy, the set of medical imaging data sampled from a plurality of humans in one or more imaging modalities, wherein the medical imaging data comprises image slices taken along at least one of coronal, sagittal and axial planes relative to the human cranial anatomy;
re-slicing at least a portion of the medical imaging data through one or more planes that are at an angular orientation with respect to at least one of the coronal, sagittal and axial planes, thereby obtaining re-sliced medical imaging data;
training a first artificial neural network (ANN) engine using a portion of the medical imaging data that has not been re-sliced and a portion of the re-sliced medical imaging data, wherein the medical imaging data is appropriately labeled, to obtain a first validated and tested ANN engine configured to distinguish one or more regions of interest (ROIs) from one or more regions of avoidance (ROAs) in a human brain image;
executing the first ANN engine, in response to an input image of a patient's brain obtained using a particular imaging modality, to identify at least one particular ROI in the patient's brain to facilitate planning of an optimal trajectory for implanting a DBS lead having one or more electrodes in the at least one particular ROI while avoiding any ROAs identified in the patient's brain;
blending two or more co-registered image slices selected from at least one of the medical imaging data that has not been re-sliced or the portion of the re-sliced medical imaging data to obtain hybrid image slices;
training a second ANN engine using a portion of the hybrid image slices to obtain a second validated and tested ANN engine configured to distinguish one or more ROIs from one or more ROAs in the human brain image; and
executing the first and second ANN engines separately with respect to the input image of the patient's brain and combining the ROI and ROA identifications obtained respectively therefrom for improving quality of identification of the at least one particular ROI.

2. The method as recited in claim 1 , wherein training the first ANN comprises
training the first ANN engine using a portion of the hybrid image slices in addition to the medical imaging data that has not been re-sliced and the portion of the re-sliced medical imaging data.

3. The method as recited in claim 1 , further comprising performing, prior to the training, morphological image processing of image slices of the medical imaging data that has not been re-sliced or the portion of the re-sliced medical imaging data, wherein the morphological image processing includes at least one of edge detection, contrast boosting and shape detection.

4. The method as recited in claim 1 , further comprising performing a dropout technique with respect to the first ANN engine wherein a select number of computational nodes are dropped from a particular neural network layer in each training epoch.

5. The method as recited in claim 1 , further comprising:
building an electrode scene with respect to the at least one particular ROI of the patient's brain image for placing the DBS lead thereat; and
determining the optimal trajectory for implanting the DBS lead in the patient's brain relative to a particular electrode of the DBS lead.

6. The method as recited in claim 5 , further comprising:
co-registering a computed tomography (CT) image of the patient's brain with the input image of the patient having the at least one particular ROI identified for stimulation, wherein the input image of the patient's brain comprises one of a pre-operative or intra-operative magnetic resonance imaging (MRI) scan; and
obtaining an entry point coordinate set and a target point coordinate set with respect to the patient's brain for performing an implant procedure to implant the DBS lead using the optimal trajectory, wherein the entry point coordinate set is operative to identify a burr hole location on the patient's cranium and the target point coordinate set is operative to identify a location relative to the at least one particular ROI in the patient's brain.

7. The method as recited in claim 6 , further comprising:
providing the entry point coordinate set, the target point coordinate set and data relating to the optimal trajectory to a stereotactic surgery system including a guiding apparatus containing the DBS lead; and
automatically guiding the DBS lead to the at least one particular ROI based on the entry point coordinate set, the target point coordinate set and the data relating to the optimal trajectory data to place the particular electrode proximate to the at least one particular ROI.

8. A computer-implemented system configured to facilitate electrode trajectory planning for deep brain stimulation (DBS), the system comprising:
one or more processors; and
a persistent memory having program instructions stored thereon, the program instructions, when executed by the one or more processors, configured to perform:
obtaining a set of medical imaging data pertaining to human cranial anatomy, the set of medical imaging data sampled from a plurality of humans in one or more imaging modalities, wherein the medical imaging data comprises image slices taken along at least one of coronal, sagittal and axial planes relative to the human cranial anatomy;
re-slicing at least a portion of the medical imaging data through one or more planes that are at an angular orientation with respect to at least one of the coronal, sagittal and axial planes, thereby obtaining re-sliced medical imaging data;
training a first artificial neural network (ANN) engine using a portion of the medical imaging data that has not been re-sliced and a portion of the re-sliced medical imaging data, wherein the medical imaging data is appropriately labeled, to generate a first validated and tested ANN engine configured to distinguish one or more regions of interest (ROIs) from one or more regions of avoidance (ROAs) in a human brain image;
in response to an input image of a patient's brain obtained using a particular imaging modality, executing the first ANN engine to identify at least one particular ROI in the patient's brain to facilitate planning of an optimal trajectory for implanting a DBS lead having one or more electrodes in the at least one particular ROI while avoiding any ROAs identified in the patient's brain;
blending two or more co-registered image slices selected from at least one of the medical imaging data that has not been re-sliced or the portion of the re-sliced medical imaging data to obtain hybrid image slices;
training a second ANN engine using a portion of the hybrid image slices to generate a second validated and tested ANN engine configured to distinguish between one or more ROIs from one or more ROAs in the human brain image; and
executing the first and second ANN engines separately with respect to the input image of the patient's brain and combining the ROI and ROA identifications obtained respectively therefrom for improving quality of identification of the at least one particular ROI.

9. The system as recited in claim 8 , wherein, to the train the first ANN engine, the program instructions further comprise instructions configured to perform:
training the first ANN engine using a portion of the hybrid image slices in addition to the medical imaging data that has not been re-sliced and the portion of the re-sliced medical imaging data.

10. The system as recited in claim 8 , wherein the program instructions further comprise instructions configured to

Assignees

• Advanced Neuromodulation Systems Inc

Inventors

• Pathak Yagna
• Zhang Simeng
• Zhu Dehan
• Kyani Anahita
• Park Hyun-Joo
• Ross Erika

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/098

  Distributed learning, e.g. federated learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/096

  Transfer learning · CPC title

• A61B34/10 Primary

  Computer-aided planning, simulation or modelling of surgical operations · CPC title