Molecular Imaging Signal Upsampling for Energy-Efficient Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing molecular imaging systems face challenges in achieving high energy resolution and image quality due to low sampling frequencies used for electrical signal processing, which result in information loss and increased energy consumption when higher frequencies are employed.
Innovation Solution
Implementing a target machine learning model in the detector to convert signals from a low sampling frequency to a higher frequency, thereby improving energy resolution and image quality while reducing energy consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a high sampling frequency is used to improve energy resolution and image quality, then measurement precision is improved, but energy consumption increases
Solution Approach 1:
The patent applies preliminary action by pre-training a machine learning model offline to learn the mapping relationship between low-frequency and high-frequency signals. During actual detection, the pre-trained model is used to convert low-frequency signals to high-frequency signals, avoiding the need for real-time high-frequency sampling and thus reducing energy consumption while maintaining measurement precision.
Solution Approach 2:
The patent replaces the mechanical/electrical high-frequency sampling system with a machine learning-based signal conversion system. Instead of physically sampling at high frequencies (which consumes energy), the system uses a trained model to computationally generate high-frequency equivalent signals from low-frequency inputs, substituting physical sampling with intelligent processing.
2Use of energy by moving object
If a low sampling frequency is used to reduce energy consumption, then energy consumption is reduced, but information loss increases
Solution Approach 1:
The patent introduces a machine learning model as an intermediary between the low-frequency sampled signal and the desired high-frequency signal. The model learns the complex mapping relationship during training and acts as a bridge to reconstruct high-frequency signal characteristics from low-frequency inputs, preventing information loss that would otherwise occur with low-frequency sampling.
Solution Approach 2:
The patent changes the parameter of sampling frequency from its physical measurement value to a learned equivalent value through machine learning. The model transforms the low-frequency signal parameters into high-frequency equivalent parameters, effectively changing the frequency parameter without requiring actual high-frequency sampling, thus preserving signal information while reducing energy consumption.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system enhances image quality by maintaining high energy resolution with reduced energy consumption, utilizing a target machine learning model to convert signals from a low to a higher sampling frequency.
Implementation Method 1
The photoelectric conversion component (e.g., a silicon photomultiplier (SiPM)) may convert the optical signals into electrical signals.
Data Source
AI summary
Methods and systems for signal processing in molecular imaging. The system may include at least one storage device including a set of instructions and at least one processor in communication with the storage device. The at least one processor may obtain a first signal that is acquired by sampling, according to a first sampling frequency, an electrical signal of a detector. The at least one processor may also generate, based on the first signal and a target machine learning model, a second signal, the second signal corresponding to a second sampling frequency that is different from the first sampling frequency. The target machine learning model may specify a target mapping between the first signal and the second signal. The at least one processor may further generate an image based on the second signal.