Neural Network Tissue Segmentation and Bounding Box Merging

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for measuring human body tissue from medical images are manual, repetitive, and prone to errors and variability among radiologists, limiting their accuracy and efficiency.

Innovation Solution

A computer-implemented method using a trained neural network to identify and segment human body tissue from medical images, compute bounding boxes, determine intersections, and merge segments to accurately measure tissue size.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual measurement by radiologists is used, then expertise-based interpretation is achieved, but measurement errors and variability between experts occur

Engineering Contradiction:
Improvemeasurement accuracyVSAvoidconsistency between experts
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The system enables self-service measurement by automatically processing medical images through neural networks and algorithms, eliminating the need for manual radiologist interpretation. The computer-implemented method independently performs segmentation, bounding box computation, and measurement calculation without human intervention, thereby removing inter-expert variability while maintaining measurement accuracy.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical human interpretation process with an automated computational system. Neural networks and algorithms substitute for radiologist expertise, performing image analysis, segmentation, and measurement tasks that were previously done manually. This substitution eliminates human error and inconsistency while maintaining or improving measurement precision.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If manual measurement processes are used, then radiologist expertise is utilized, but the process becomes repetitive and tedious

Engineering Contradiction:
Improvemeasurement qualityVSAvoidmeasurement efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The system performs self-service by automatically executing the complete measurement workflow from image input to final measurement output. The computer-implemented method handles segmentation, bounding box computation, intersection determination, and measurement calculation without requiring radiologist intervention, thereby eliminating repetitive manual tasks while maintaining measurement quality.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary actions by pre-processing images through neural network segmentation and pre-computing bounding boxes before final measurement. This automated preparation work eliminates the need for radiologists to perform repetitive preliminary analysis, freeing them from tedious tasks while ensuring consistent quality through algorithmic precision.

Inventive Principle:
Principle #10Preliminary action

3Productivity

If automated approaches like maskrcnn are used, then productivity is improved, but specialized models per organ are required increasing complexity

Engineering Contradiction:
Improveautomation speedVSAvoidmodel specialization requirements
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent implements a universal measurement framework that can process multiple organ types and imaging modalities through a single computer-implemented method. The system uses general-purpose neural networks and algorithms that adapt to different organs (liver, kidney, spleen, etc.) and image types (CT, MRI, ultrasound) without requiring separate specialized models, thereby maintaining high productivity while reducing system complexity.

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

Solution Approach 2:

The system applies segmentation to divide the complex task of multi-organ measurement into distinct processing stages: image segmentation, bounding box computation, intersection determination, and measurement calculation. This segmented approach allows a single universal model to handle diverse organs efficiently by processing each through the same standardized pipeline, reducing the need for organ-specific specialized models.

Inventive Principle:
Principle #1Segmentation

4Measurement precision

If radiologists manually select best practice images, then measurement accuracy is maintained, but time consumption increases

Engineering Contradiction:
Improvemeasurement accuracyVSAvoidimage selection time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs self-service by automatically selecting and processing optimal images without radiologist intervention. The computer-implemented method independently evaluates multiple images, determines the best practice image for measurement, and proceeds with automated analysis, thereby eliminating the time radiologists spend on manual image selection while maintaining measurement accuracy through algorithmic optimization.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary image evaluation and selection automatically before the measurement process. Neural networks pre-analyze multiple images to identify the optimal image for measurement based on quality metrics and anatomical clarity, eliminating the need for radiologists to manually review and select images, thereby saving time while ensuring measurement accuracy is maintained.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4571642A1Measurement of human body tissue
Publication Date: 2025.06.18 DASSAULT SYSTEMES SA
  • EP4571642A1 patent drawingFigure 1
  • EP4571642A1 patent drawingFigure 2
  • EP4571642A1 patent drawingFigure 3

AI summary

The disclosure notably relates to a computer-implemented method for measuring a human body tissue from a set of medical images representing the human body tissue. The method comprises obtaining a trained neural network configured for outputting segments of human body tissue. The method also comprises applying the trained neural network to the set. The method thereby identifies one or more segments of human body tissue for at least two images of the set. The method also computes a bounding box enclosing each segment. The method also determines an intersection between a pair of bounding boxes. If the intersection of the pair is non-empty, the method determines an intersection between the segments. If the intersection between the segments is non-empty, the method merges the segments by computing a resulting bounding box enclosing the segments. The method also measures the size of the segments comprised in the resulting bounding box.