Fat Prediction Model for CT Image Fat Content Measurement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for measuring fat content, such as ultrasonic waves and impedance methods, struggle to accurately quantify fat content in specific regions of the body, like the liver, and cannot be applied to computed tomography (CT) images.
Innovation Solution
A method and apparatus using a computed tomography (CT) image to measure fat content by training a fat prediction model with learning data, including CT images or noise images, to generate a fat distribution image, which is then used for accurate fat content measurement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ultrasonic waves or impedance methods are used to measure fat content, then the measurement can be performed, but the accuracy of fat content quantification in specific regions (e.g., liver) is insufficient
Solution Approach 1:
The patent introduces a fat prediction model trained on CT images as an intermediary to bridge the gap between CT imaging and fat content quantification. This model acts as a mediator that translates CT image data into accurate fat distribution maps, enabling both high measurement precision and adaptability to specific tissue regions like the liver.
Solution Approach 2:
The patent transforms the measurement approach by changing from direct physical measurement (ultrasonic waves, impedance) to a computational parameter transformation approach. CT images are processed through a trained prediction model that outputs fat distribution parameters, achieving accurate regional fat quantification without the limitations of traditional methods.
2Measurement precision
If PDFF method using MRI is used to accurately quantify fat content, then the fat distribution can be measured, but the method cannot be applied to CT images
Solution Approach 1:
The patent creates a computational copy of the PDFF methodology that works with CT images instead of MRI. By training a prediction model on CT images with corresponding fat distribution labels, the system replicates the accurate fat measurement capability of PDFF but adapts it to the CT imaging modality, making it compatible with widely available CT scanners.
Solution Approach 2:
The patent replaces the physical MRI-based PDFF measurement mechanism with a computational approach using CT images and machine learning. Instead of relying on MRI's specific physical properties for fat quantification, the system uses a trained neural network to infer fat distribution from CT attenuation values, substituting one measurement mechanism for another that is more universally applicable.
3Measurement precision
If a fat prediction model is trained using learning data, then accurate fat distribution images can be generated from CT images, but additional training and processing time is required
Solution Approach 1:
The patent performs the computationally intensive model training in advance during a preliminary phase. Once the fat prediction model is trained on labeled CT images, it can be deployed for rapid inference on new patient scans. This preliminary action separates the time-consuming training phase from the clinical application phase, reducing the time loss during actual patient diagnostics.
Data Source
AI summary
Provided is a method and apparatus for measuring fat content using a computed tomography (CT) image. The fat content measurement apparatus trains a fat prediction model by using learning data including a CT image or noise image for learning and generates a fat distribution image to be used for fat content measurement by the fat prediction model having completed learning upon receiving a CT image for diagnosis.