Multiphasic Medical Image Classification Using Phase-Aware Neural Networks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The complexity of medical imaging reporting frameworks has made them less feasible in high-volume practices, necessitating a need for computational decision-support tools to improve workflow efficiency in radiological diagnosis, particularly for classifying features in multi-phasic medical images like liver cancer.

Innovation Solution

A computer-implemented method using neural network-based learning algorithms combines medical image data with phase identifiers to enhance classification accuracy, leveraging different phases of multiphasic images for improved training and classification, employing convolutional neural networks and the Liver Imaging Reporting and Data System (LI-RADS) framework.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If standardized image analysis frameworks (e.g., LI-RADS) are implemented to improve radiological diagnosis, then measurement precision and reliability are improved, but device complexity and ease of operation deteriorate

Engineering Contradiction:
Improvediagnosis accuracyVSAvoidframework complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces manual radiological assessment (mechanical/human process) with an automated machine learning system that processes multi-phasic medical images. The system automatically extracts features, determines enhancement patterns, and classifies lesions according to LI-RADS criteria, substituting the complex manual framework with an automated computational system that maintains diagnostic accuracy while reducing operational complexity.

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

Solution Approach 2:

The machine learning system performs self-service by automatically executing the entire LI-RADS classification workflow without requiring manual intervention at each step. The system independently processes multi-phasic images, extracts relevant features, determines arterial enhancement and washout patterns, and generates classification results, enabling the framework to serve itself rather than requiring continuous human operation.

Inventive Principle:
Principle #25Self-service

2Productivity

If manual decision-making is used for image feature classification, then ease of operation is maintained, but productivity and measurement precision worsen

Engineering Contradiction:
Improveworkflow efficiencyVSAvoidclassification consistency
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent replaces manual classification processes with an automated machine learning system that consistently applies LI-RADS criteria across all cases. The system processes multiple images rapidly without fatigue or variation in judgment, simultaneously improving productivity through automated batch processing and measurement precision through consistent application of classification rules.

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

Solution Approach 2:

The system incorporates feedback mechanisms where the machine learning model is trained on labeled data with known outcomes, allowing it to learn from correct classifications and improve its performance. The system receives feedback during training and can be refined to enhance both speed and accuracy of classification decisions.

Inventive Principle:
Principle #23Feedback

3Measurement precision

If multiple phases of multiphasic images are analyzed to improve classification accuracy, then measurement precision is improved, but device complexity and loss of time worsen

Engineering Contradiction:
Improveclassification accuracyVSAvoidanalysis time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary action by pre-processing and analyzing all phases of multiphasic images simultaneously during the initial processing stage. Rather than sequentially examining each phase, the machine learning system processes arterial phase, portal venous phase, and delayed phase images in parallel, extracting features and determining enhancement patterns across all phases before final classification, thereby reducing total analysis time while maintaining high accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent merges the analysis of multiple image phases into a unified processing workflow. The machine learning system combines features from arterial phase, portal venous phase, and delayed phase images into a comprehensive analysis, determining enhancement patterns and washout characteristics by integrating information across all phases simultaneously rather than separately.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP4252180B1Image feature classification
Publication Date: 2026.03.18 KONINKLIJKE PHILIPS NV
  • EP4252180B1 patent drawingFigure 1
  • EP4252180B1 patent drawingFigure 2
  • EP4252180B1 patent drawingFigure 3

AI summary

A method and system for image feature classification using a NN-based learning algorithms to make a decision about a feature in a medical image or image part. In particular, embodiments may make use of a phase of a multi-phasic image to improve classification accuracy. For instance, embodiments may combine different phases of multiphasic images as training data.