SPECT Imaging Prediction Model Using Deep Convolutional Neural Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
SPECT imaging faces challenges with long imaging times, causing radiation damage and resulting in low signal-to-noise ratio and motion artifacts, which compromise image quality.
Innovation Solution
A SPECT imaging prediction model creation method using a deep convolutional neural network is developed, training on pairs of standard and short acquisition duration images to predict high-quality images within a shorter acquisition time, incorporating CT images for improved training and reducing motion artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If standard acquisition duration SPECT imaging is used, then image quality is maintained, but imaging time is long and radiation damage occurs
Solution Approach 1:
The patent applies preliminary action by training a deep convolutional neural network model in advance using pairs of standard and short acquisition duration SPECT images. The model learns the transformation relationship between short and standard duration images during the training phase, enabling rapid prediction of high-quality images from short acquisition data without requiring actual standard duration scanning for each patient
Solution Approach 2:
The patent uses copying by creating a synthetic training dataset where short acquisition duration SPECT images are paired with their corresponding standard acquisition duration SPECT images. The neural network learns to copy the quality characteristics of standard duration images onto short duration images through the training process, enabling the generation of high-quality images from rapid scans
2Loss of time
If imaging angle is reduced to shorten imaging time, then imaging duration decreases, but signal-to-noise ratio deteriorates and artifacts increase
Solution Approach 1:
The patent applies parameter changes by modifying the acquisition parameters to use short acquisition duration scanning with a reduced number of projection angles. The neural network model compensates for the reduced angular sampling by learning the mapping relationship between short and standard duration images, maintaining image quality despite the changed acquisition parameters
Data Source
AI summary
A SPECT imaging prediction model creation method, apparatus, device, and storage medium. The method includes: obtaining a training set including a plurality of scanning image groups, wherein each scanning image group includes a standard acquisition duration SPECT image and a short acquisition duration SPECT image that corresponding to each other; performing network construction on the basis of deep convolutional neural network to obtain a network to be trained; taking the short acquisition duration SPECT image in the training set as input-side training data, taking the standard acquisition duration SPECT image in the training set as output-side training data, training the network to be trained to obtain a SPECT imaging prediction model, so as to predict SPECT prediction image of short acquisition duration SPECT image under standard acquisition duration by using the SPECT imaging prediction model. The SPECT imaging time is significantly reduced while maintaining the imaging quality of medical images.