Facial Action-Based Patient Motion Prediction for Medical Imaging

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Patient movements during medical imaging examinations, such as CT or MRI, cause image blurs and artifacts, often requiring repeat scans, which are time-consuming and expose patients to unnecessary radiation.

Innovation Solution

A computer-implemented method using a trained machine learning model to generate patient motion data from facial actions, predicting non-facial body part movements based on facial expressions, allowing real-time adjustment of imaging protocols to minimize artifacts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If motion correction algorithms are applied to reduce artifacts, then image quality is improved, but technical limitations prevent complete elimination of motion artifacts

Engineering Contradiction:
Improveimage qualityVSAvoidartifact reduction effectiveness
Core Design Contradiction:
Manufacturing precisionVSReliability

Solution Approach 1:

The system performs preliminary detection of patient motion through facial action analysis before the motion affects the imaging quality. By detecting facial movements that indicate potential body part motion, the system can alert operators or adjust imaging parameters in advance, preventing artifact formation rather than attempting to correct it afterward.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system uses facial action as an intermediary indicator to predict and monitor patient motion. Instead of directly measuring body part movement, the system detects facial expressions and movements that correlate with overall patient comfort and motion tendencies, using this intermediate data to infer and address potential imaging artifacts.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If repeat scans are performed to ensure image quality, then diagnostic reliability is improved, but examination time and radiation exposure increase

Engineering Contradiction:
Improvediagnostic reliabilityVSAvoidexamination time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system continuously monitors patient facial actions during the imaging examination and provides real-time feedback about motion detection. This feedback loop allows operators to adjust patient positioning, communicate with anxious patients, or pause the examination to address motion issues, thereby preventing the need for repeat scans and reducing total examination time and radiation exposure.

Inventive Principle:
Principle #23Feedback

3Ease of operation

If operators monitor patient comfort through oral communication, then patient compliance can be assessed, but time consumption and effort increase

Engineering Contradiction:
Improvepatient compliance assessmentVSAvoidcommunication time
Core Design Contradiction:
Ease of operationVSLoss of time

Solution Approach 1:

The system enables automatic monitoring of patient comfort and motion without requiring continuous operator intervention. The machine learning model automatically analyzes facial action data from cameras, detecting signs of pain, anxiety, or discomfort, and generates motion predictions without operator involvement, freeing operators to focus on critical tasks while maintaining patient monitoring.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP4636778A1A method for generating patient motion data for a medical imaging examination of a patient
Publication Date: 2025.10.22 SIEMENS HEALTHINEERS AG
  • EP4636778A1 patent drawingFigure 1
  • EP4636778A1 patent drawingFigure 2~3
  • EP4636778A1 patent drawingFigure 4~5

AI summary

The invention relates in one aspect to a computer-implemented method for generating patient motion data for a medical imaging examination of a patient, the method comprising: - receiving input data, the input data comprising information regarding a facial action of the patient, - applying a trained machine learning model to the input data, wherein the patient motion data is generated based on the input data, the patient motion data comprising information regarding a motion of a non-facial body part of the patient, wherein the information regarding the motion of the non-facial body part of the patient is generated based on the information regarding the facial action of the patient.