Left atrium shape reconstruction from sparse location measurements using neural networks

A system, comprising:

• a memory, which is configured to store example representations of geometrical shapes of a given type of organ;
• and
• a processor, which is configured to:
• upload the example representations from the memory;
• in a training phase, train a neural network model using the example representations to produce a smooth reconstruction, wherein the neural network model comprises a regularization function comprising smoothing spatial weights, and store the trained neural network model in the memory;
• and
• in a modeling phase, apply the trained neural network model to a set of location measurements acquired in an organ of the given type.

A method, comprising:

• in a processor, receiving example representations of geometrical shapes of a given type of organ;
• in a training phase, training a neural network model using the example representations to produce a smooth reconstruction by minimizing a regularization-function that comprises derivatives of weights of a first layer only of the neural-network model;
• and
• in a modeling phase, applying the trained neural network model to a set of location measurements acquired in an organ of the given type, to produce a three-dimensional model of the organ.

A system, comprising:

• a memory, which is configured to store example representations of geometrical shapes of a given type of organ;
• and
• a processor, which is configured to:
• upload the example representations from the memory;
• in a training phase, train a neural network model using the example representations to produce a smooth reconstruction by minimizing a regularization-function that comprises derivatives of weights of a first layer only of the neural-network model;
• and
• in a modeling phase, applying the trained neural network model to a set of location measurements acquired in an organ of the given type, to produce a three-dimensional model of the organ.