Systems and methods for electrode orientation determination in deep brain stimulation (dbs)

A computer-implemented method for estimating deep brain stimulation (DBS) lead orientation, the method comprising:

• generating an initial image dataset from a plurality of clinical images and a plurality of phantom images, each image in the initial image dataset including a depiction of an implanted DBS lead and a label of an orientation of the implanted DBS lead;
• modifying the initial image dataset by duplicating and modifying at least a subset of the initial image dataset, the modifying including at least one of down-sampling a respective image or adding noise to a respective image of the subset of the initial image dataset, to generate a modified image dataset;
• modifying the modified image dataset by duplicating and re-slicing at least a subset of the modified image dataset, the re-slicing including re-slicing a respective image along an alternative primary imaging axis, to generate an integrated image dataset;
• partitioning the integrated image dataset into a preliminary training image dataset and a testing image dataset;
• modifying the preliminary training image dataset by duplicating and re-sizing at least a subset of the preliminary training image dataset with a localized field of view around the respective depiction of the implanted DBS lead, to generate a training image dataset;
• training a machine-learning model using the training image dataset, including training the machine-learning model to associate one or more image artifacts with the orientation of the implanted DBS lead;
• and
• executing the trained machine-learning model to estimate, during a DBS implantation procedure, an orientation of a subject implanted DBS lead.

Non-transitory computer-readable media having computer-executable instructions thereon, wherein when executed by a processor of a computing device communicatively coupled to a memory device, cause the processor of the computing device to:

• generate an initial image dataset from a plurality of clinical images and a plurality of phantom images, each image in the initial image dataset including a depiction of an implanted DBS lead and a label of an orientation of the implanted DBS lead;
• modify the initial image dataset by duplicating and modifying at least a subset of the initial image dataset, the modifying including at least one of down-sampling a respective image or adding noise to a respective image of the subset of the initial image dataset, to generate a modified image dataset;
• modify the modified image dataset by duplicating and re-slicing at least a subset of the modified image dataset, the re-slicing including re-slicing a respective image along an alternative primary imaging axis, to generate an integrated image dataset;
• partition the integrated image dataset into a preliminary training image dataset and a testing image dataset;
• modify the preliminary training image dataset by duplicating and re-sizing at least a subset of the preliminary training image dataset with a localized field of view around the respective depiction of the implanted DBS lead, to generate a training image dataset;
• train a machine-learning model using the training image dataset, including training the machine-learning model to associate one or more image artifacts with the orientation of the implanted DBS lead;
• and
• execute the trained machine-learning model to estimate, during a DBS implantation procedure, an orientation of a subject implanted DBS lead.