Variable-Confidence Machine Learning for Medical Image Regression

Solution Overview

Problem

Existing machine learning models for medical image regression tasks require separate training for each desired confidence quantile, leading to excessive computing resource consumption and inefficiency.

Innovation Solution

A machine learning model is configured to receive a medical image and a user-specified confidence indicator as input, allowing it to generate regression task intervals based on variable confidence quantiles without retraining.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a separate machine learning model is trained for each unique confidence quantile, then the machine learning model can provide accurate regression task intervals for different confidence levels, but the computing resource consumption increases excessively

Engineering Contradiction:

Improveaccuracy of regression task intervalsVSAvoidcomputing resource consumption

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The machine learning model is designed to perform multiple functions by accepting variable confidence quantiles as input. A single model can generate regression task intervals for different confidence levels (e.g., 90%, 95%, 99%) without requiring separate models for each quantile, thus achieving multi-functionality while reducing computing resource consumption

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The model utilizes parameter changes by incorporating confidence quantile as a variable input parameter. Instead of training separate models for different confidence levels, the system changes the confidence quantile parameter dynamically during inference, allowing one model to adapt to different confidence requirements without retraining

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If multiple separate machine learning models are trained for different confidence quantiles, then variable confidence regression can be achieved, but the device complexity and memory usage increase

Engineering Contradiction:

Improvevariable confidence regression capabilityVSAvoidnumber of models required

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The machine learning model is designed to perform multiple functions by accepting variable confidence quantiles as input. A single model can generate regression task intervals for different confidence levels (e.g., 90%, 95%, 99%) without requiring separate models for each quantile, thus achieving multi-functionality while reducing computing resource consumption

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system introduces dynamic adaptability by allowing the confidence quantile to be variable during inference. The model can adjust its output based on the input confidence parameter, enabling dynamic adaptation to different confidence requirements without maintaining multiple static models

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS12444172B2Variable confidence machine learning

Publication Date: 2025.10.14 GE PRECISION HEALTHCARE LLC

AI summary

Systems/techniques that facilitate variable confidence machine learning are provided. In various embodiments, a system can access a medical image. In various aspects, the system can perform, via execution of a machine learning model, a regression task on the medical image, wherein the machine learning model can receive as input both the medical image and a user-specified confidence indicator.