Medical Image Autoencoder Compression Using an Extra Dimension

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for compressing digital medical image datasets, such as those from MRI scanners, are inefficient in reducing data redundancy, leading to high storage requirements and resource-intensive data transmission.

Innovation Solution

Utilizing an autoencoder network with an additional dimension to exploit data redundancy in medical image datasets, specifically designed for medical scanners like MRI, CT, PET, and SPECT, to compress and reconstruct images effectively, reducing storage needs and optimizing transmission.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If conventional compression methods are used on medical image data, then storage space is reduced, but data redundancy is not sufficiently exploited leading to inefficient compression

Engineering Contradiction:
Improvestorage spaceVSAvoidcompression efficiency
Core Design Contradiction:
Quantity of substanceVSProductivity

Solution Approach 1:

The patent introduces an additional dimension to the autoencoder network architecture specifically designed to exploit data redundancy in medical image datasets. This dimensional extension allows the network to process and compress redundant information more effectively across multiple data dimensions (spatial, temporal, contrast, coil), achieving superior compression ratios while maintaining image quality.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The invention modifies the network architecture parameters by adding an extra dimension to the autoencoder, enabling it to adapt to different medical imaging modalities and redundancy patterns. This parameter change transforms the network's capability to handle various types of medical image data (MRI, CT, PET, SPECT) with different redundancy characteristics.

Inventive Principle:
Principle #35Parameter changes

2Productivity

If data redundancy is not exploited, then transmission resources are consumed, but compression is insufficient

Engineering Contradiction:
Improvetransmission efficiencyVSAvoiddata volume
Core Design Contradiction:
ProductivityVSQuantity of substance

Solution Approach 1:

By extending the autoencoder with an additional dimension, the system efficiently identifies and compresses redundant data patterns across multiple dimensions including temporal, spatial, contrast, and coil dimensions. This enables significant reduction in data volume for transmission while maintaining essential diagnostic information.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Adaptability or versatility

If autoencoder network is designed without additional dimension, then architecture is simple, but data redundancy exploitation is limited

Engineering Contradiction:
Improveredundancy exploitation capabilityVSAvoidnetwork architecture
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent adds one extra dimension to the standard autoencoder architecture, creating a more versatile network capable of exploiting various types of data redundancy in medical images. This controlled increase in complexity significantly enhances the network's adaptability to different medical imaging scenarios and redundancy patterns.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The extended autoencoder network is designed to handle multiple types of medical image data (MRI, CT, PET, SPECT) and various redundancy dimensions (temporal, spatial, contrast, coil) through a single unified architecture, making it universally applicable across different medical imaging applications.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4621793A1Technology for compressing medical image data
Publication Date: 2025.09.24 SIEMENS HEALTHINEERS AG
  • EP4621793A1 patent drawingFigure 1~2
  • EP4621793A1 patent drawingFigure 3
  • EP4621793A1 patent drawingFigure 4

AI summary

The invention relates to a technique for compressing a digital medical image dataset obtained by means of a medical scanner. A computer-implemented method (100) comprises a step of receiving (S104), at an input layer of a trained autoencoder, the digital medical image dataset from the medical scanner. The trained autoencoder comprises at least one extra dimension in the input layer and optionally in an output layer. The at least one extra dimension is provided in particular in response to a received (S102) indication of at least one data redundancy dimension of the received digital medical image dataset.The method (100) further comprises a step of compressing (S106), by means of the trained autoencoder, at least a portion of the received (S104) digital medical image data set using the received (S102) indication of the at least one data redundancy dimension and/or data received in the at least one extra dimension.