Early Dynamic PET Prediction for Faster Disease-Specific Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
PET image-based diagnosis in nuclear medicine is performed at a relatively late time point after tracer injection, requiring an improved method for earlier prediction of anatomical and disease-specific information.
Innovation Solution
A method involving early dynamic image data processing using first and second prediction models to predict anatomical and disease-specific information earlier than conventional PET imaging, utilizing a computing device with processors to learn and process PET image data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If PET image-based diagnosis is performed at a late time point after tracer injection, then accurate diagnostic information can be obtained, but the diagnostic time is delayed
Solution Approach 1:
The patent applies preliminary action by performing PET imaging at an early time point (e.g., 5-30 minutes) after tracer injection, before the conventional late time point (1.5-3 hours). This early imaging captures the tracer distribution during the initial phase, and prediction models are then used to infer the anatomical and disease-specific information that would typically require waiting for the late time point, thereby advancing the diagnostic process without sacrificing accuracy
Solution Approach 2:
The patent uses prediction models (such as deep learning models) to create a copy or surrogate of the late-time-point diagnostic information based on early-time-point imaging data. The model learns the relationship between early and late time point images from training data, then generates predicted late-time-point images from early images, effectively copying the diagnostic information that would otherwise require waiting to obtain
2Loss of time
If early dynamic image data is used for prediction, then diagnostic time is reduced, but the complexity of the prediction model increases
Solution Approach 1:
The patent introduces prediction models as an intermediary between the early dynamic image data and the final diagnostic information. Instead of directly analyzing early images for diagnostic purposes, the model acts as a mediator that processes the early images, learns the underlying patterns and relationships, and generates the equivalent of late-time-point diagnostic information, thereby bridging the gap between early imaging and accurate diagnosis
Solution Approach 2:
The patent transforms the early dynamic image data through various parameter changes and processing steps in the prediction model, including spatial normalization, intensity normalization, and feature extraction. These parameter transformations convert the raw early images into a format suitable for accurate prediction, managing the complexity by systematically processing the data through defined computational steps
Data Source
AI summary
Provided are a dynamic image data-based object state prediction method and a computer device performing the same. The computing device includes a memory and at least one processor communicating with the memory. The processor obtains early dynamic image data corresponding to an early section after a medicine is injected into a learning object until a preset time point, learns a first prediction model for predicting first image data indicating anatomical information about the learning object corresponding to a time point earlier than a first section by using, as an input, early dynamic image data corresponding to the first section of the early section, and learns a second prediction model for predicting second image data indicating disease-specific information about the learning object corresponding to a reference time point after the early section by using, as an input, early dynamic image data corresponding to a second section of the early section.