# US 2024/0203039 A1 · Interpretable task-specific dimensionality reduction

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2024/0203039 A1
- Document ID: US20240203039A1
- Status: Published application
- Kind: A1
- Application number: 18/065,964
- Filed: 2022-12-14
- Published: 2024-06-20
- Earliest claimed: 2022-12-14
- Assignees: GE Precision Healthcare LLC
- Applicants: GE Precision Healthcare LLC
- Inventors: Deepa Anand; Bipul Das; Vanika Singhal; Rakesh Mullick; Sanjay Kumar NT
- Web page: https://patentgrep.com/patent/US20240203039A1

## CPC

- G06T 15/20: Perspective computation
- G06T 15/08: Volume rendering
- G06V 10/82: using neural networks
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

Systems/techniques that facilitate interpretable task-specific dimensionality-reduction are provided. In various embodiments, a system can access a three-dimensional medical image. In various aspects, the system can generate, via execution of a first deep learning neural network, a voxel-wise weight map corresponding to the three-dimensional medical image and a set of projection vectors corresponding to the three-dimensional medical image. In various instances, the system can generate a set of two-dimensional projection images of the three-dimensional medical image, based on the voxel-wise weight map and the set of projection vectors. In various cases, the first deep learning neural network can be trained in a serial pipeline with a second deep learning neural network that is configured to perform an inferencing task on two-dimensional inputs. This can cause the set of two-dimensional projection images to be tailored to the inferencing task.

## Claims

### Claim 1

A system, comprising:
a processor that executes computer-executable components stored in a non-transitory computer-readable memory, the computer-executable components comprising:
an access component that accesses a three-dimensional medical image;
a parameter component that generates, via execution of a first deep learning neural network, a voxel-wise weight map corresponding to the three-dimensional medical image and a set of projection vectors corresponding to the three-dimensional medical image; and
a projection component that generates a set of two-dimensional projection images of the three-dimensional medical image, based on the voxel-wise weight map and the set of projection vectors.

### Claim 2 (depends on 1)

The system of claim 1, wherein the first deep learning neural network is trained in a serial pipeline with a second deep learning neural network that is configured to perform an inferencing task on two-dimensional inputs, such that the set of two-dimensional projection images are tailored to the inferencing task.

### Claim 3 (depends on 2)

The system of claim 2, wherein the computer-executable components further comprise:
a training component that executes the second deep learning neural network on the set of two-dimensional projection images, thereby yielding an inferencing task output.

### Claim 4 (depends on 3)

The system of claim 3, wherein the training component updates internal parameters of the first deep learning neural network based on an error between the inferencing task output and a ground-truth annotation corresponding to the three-dimensional medical image, or between the inferencing task output and a projection of the ground-truth annotation.

### Claim 5 (depends on 4)

The system of claim 4, wherein the training component updates internal parameters of the second deep learning neural network based on the error.

### Claim 6 (depends on 1)

The system of claim 1, wherein the projection component generates the set of two-dimensional projection images by multiplying, in element-wise fashion, the voxel-wise weight map with the three-dimensional medical image, thereby yielding an altered three-dimensional medical image, and by projecting the altered three-dimensional medical image along the set of projection vectors.

### Claim 7 (depends on 1)

The system of claim 1, wherein the computer-executable components further comprise:
a display component that renders the set of two-dimensional projection images on an electronic display.

### Claim 8 (depends on 1)

The system of claim 1, wherein the set of two-dimensional projection images includes an oblique projection.

### Claim 9

A computer-implemented method, comprising:
accessing, by a device operatively coupled to a processor, a three-dimensional medical image;
generating, by the device and via execution of a first deep learning neural network, a voxel-wise weight map corresponding to the three-dimensional medical image and a set of projection vectors corresponding to the three-dimensional medical image; and
generating, by the device, a set of two-dimensional projection images of the three-dimensional medical image, based on the voxel-wise weight map and the set of projection vectors.

### Claim 10 (depends on 9)

The computer-implemented method of claim 9, wherein the first deep learning neural network is trained in a serial pipeline with a second deep learning neural network that is configured to perform an inferencing task on two-dimensional inputs, such that the set of two-dimensional projection images are tailored to the inferencing task.

### Claim 11 (depends on 10)

The computer-implemented method of claim 10, further comprising:
executing, by the device, the second deep learning neural network on the set of two-dimensional projection images, thereby yielding an inferencing task output.

### Claim 12 (depends on 11)

The computer-implemented method of claim 11, further comprising:
updating, by the device, internal parameters of the first deep learning neural network based on an error between the inferencing task output and a ground-truth annotation corresponding to the three-dimensional medical image, or between the inferencing task output and a projection of the ground-truth annotation.

### Claim 13 (depends on 12)

The computer-implemented method of claim 12, further comprising:
updating, by the device, internal parameters of the second deep learning neural network based on the error.

### Claim 14 (depends on 9)

The computer-implemented method of claim 9, wherein the generating the set of two-dimensional projection images comprises:
multiplying, by the device and in element-wise fashion, the voxel-wise weight map with the three-dimensional medical image, thereby yielding an altered three-dimensional medical image; and
projecting, by the device, the altered three-dimensional medical image along the set of projection vectors.

### Claim 15 (depends on 9)

The computer-implemented method of claim 9, further comprising:
rendering, by the device, the set of two-dimensional projection images on an electronic display.

### Claim 16 (depends on 9)

The computer-implemented method of claim 9, wherein the set of two-dimensional projection images includes an oblique projection.

### Claim 17

A computer program product for facilitating interpretable task-specific dimensionality reduction, the computer program product comprising a non-transitory computer-readable memory having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:
access a voxel array;
execute a first deep learning neural network on the voxel array, thereby yielding a voxel-wise weight map and at least one projection vector;
multiply, in element-wise fashion, the voxel array by the voxel-wise weight map, thereby yielding an altered voxel array;
project the altered voxel array along the at least one projection vector, thereby yielding at least one projected pixel array; and
render the at least one projected pixel array on an electronic display.

### Claim 18 (depends on 17)

The computer program product of claim 17, wherein the first deep learning neural network is trained in a serial pipeline with a second deep learning neural network that is configured to perform an inferencing task on two-dimensional inputs, such that the at least one projected pixel array is specific to the inferencing task.

### Claim 19 (depends on 18)

The computer program product of claim 18, wherein the program instructions are further executable to cause the processor to:
execute the second deep learning neural network on the at least one projected pixel array, thereby yielding an inferencing task output.

### Claim 20 (depends on 19)

The computer program product of claim 19, wherein the program instructions are further executable to cause the processor to:
update internal parameters of the first deep learning neural network based on an error between the inferencing task output and a ground-truth annotation corresponding to the voxel array, or between the inferencing task output and a projection of the ground-truth annotation.
