Generating implicit neural representations of medical images

A computer implemented method for using a system to generate an Implicit Neural Representation, INR, of a 3-dimensional, 3D, medical image, wherein the system comprises a first probabilistic Machine Learning (ML) module, a second probabilistic ML module, and a probabilistic Neural Radiance Field (NeRF) module, the method comprising:

• obtaining a context set of ray paths and corresponding 2-dimensional (2D) projections of the image;
• generating a plurality of context geometric bases using the first probabilistic ML module and the context set;
• generating prior distributions of a hierarchical set of latent variables using the second probabilistic ML module, the context geometric bases, and points sampled along ray paths sampled from the 3D medical image;
• and
• modulating the probabilistic NeRF module using values sampled from the prior distributions of the hierarchical set of latent variables.

A non-transitory computer-readable medium with instructions stored thereon, the instructions, when executed by a processor of a system comprising a first probabilistic Machine Learning (ML) module, a second probabilistic ML module, and a probabilistic Neural Radiance Field (NeRF) module, cause the processor to perform operations comprising:

• obtaining a training dataset comprising, for individual training 3D medical images, a context set of ray paths and corresponding 2-dimensional (2D) projections of the image, and a target set of ray paths and corresponding 2D projections of the image, the target set comprising a greater number of ray paths and corresponding 2D projections of the image than the context set;
• for individual training 3D medical images:
• generating a plurality of context geometric bases using the first probabilistic ML module and the context set;
• generating a plurality of target geometric bases using the first probabilistic ML module and the target set;
• generating prior distributions of a set of latent variables using the second probabilistic ML module, the context geometric bases, and points sampled along ray paths sampled from the training 3D medical image;
• and
• generating posterior distributions of the set of latent variables using the second probabilistic ML module, the target geometric bases, and the points sampled along ray paths sampled from the training 3D medical image;
• modulating the probabilistic NeRF module using values sampled from the prior distributions of the set of latent variables;
• using the modulated probabilistic NeRF module to predict attenuation coefficient values for points sampled along ray paths sampled from the training 3D medical image;
• and
• updating trainable parameters of the first probabilistic ML module, second probabilistic ML module, and probabilistic NeRF module to minimize an objective function.

A training node for training a system to generate an Implicit Neural Representation (INR) of a 3-dimensional (3D) medical image, wherein the system comprises a first probabilistic Machine Learning (ML) module, a second probabilistic ML module, and a probabilistic Neural Radiance Field (NeRF) module, the training node comprising processing circuitry configured to cause the training node to:

• obtain a training dataset comprising, for individual training 3D medical images, a context set of ray paths and corresponding 2-dimensional, 2D, projections of the image, and a target set of ray paths and corresponding 2D projections of the image, the target set comprising a greater number of ray paths and corresponding 2D projections of the image than the context set;
• for individual training 3D medical images:
• generate a plurality of context geometric bases using the first probabilistic ML module and the context set;
• generate a plurality of target geometric bases using the first probabilistic ML module and the target set;
• generate prior distributions of a set of latent variables using the second probabilistic ML module, the context geometric bases, and points sampled along ray paths sampled from the training 3D medical image;
• and
• generate posterior distributions of the set of latent variables using the second probabilistic ML module, the target geometric bases, and the points sampled along ray paths sampled from the training 3D medical image;
• modulate the probabilistic NeRF module using values sampled from the prior distributions of the set of latent variables;
• use the modulated probabilistic NeRF module to predict attenuation coefficient values for points sampled along ray paths sampled from the training 3D medical image;
• and
• update trainable parameters of the first probabilistic ML module, second probabilistic ML module, and probabilistic NeRF module to minimize an objective function.

An image processing node for using a system to generate an Implicit Neural Representation (INR) of a 3-dimensional (3D) medical image, wherein the system comprises a first probabilistic Machine Learning (ML) module, a second probabilistic ML module, and a probabilistic Neural Radiance Field (NeRF) module, the image processing node comprising processing circuitry configured to cause the image processing node to:

• obtain a context set of ray paths and corresponding 2-dimensional, 2D, projections of the image;
• generate a plurality of context geometric bases using the first probabilistic ML module and the context set;
• generate prior distributions of a hierarchical set of latent variables using the second probabilistic ML module, the context geometric bases, and points sampled along ray paths sampled from the 3D medical image;
• and
• modulate the probabilistic NeRF module using values sampled from the prior distributions of the hierarchical set of latent variables.