Semiconductor radiation sensor device and method of training an artificial neural network for signal processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional readout systems for photon-by-photon detection and measurement of X-ray and gamma-ray radiation are limited by slow accuracy and require high computational power, especially for real-time analysis, which is cumbersome and inefficient.
Innovation Solution
A semiconductor radiation sensor device equipped with a processing unit that includes an artificial neural network (ANN) trained on the physical and geometrical properties of semiconductor sensors, enabling real-time estimation of interaction time and three-dimensional interaction position by processing signals from a plurality of semiconductor sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional readout systems are used for photon-by-photon detection, then measurement precision is maintained, but productivity is reduced due to slow processing speed
Solution Approach 1:
The patent replaces conventional mechanical/electronic signal processing systems with an artificial neural network-based system. The ANN processes detector signals in real-time, achieving both high measurement precision for interaction position and energy, and high processing speed for real-time radiation imaging, thereby resolving the contradiction between accuracy and processing speed.
2Productivity
If high capacity computational power is used for real-time analysis, then productivity is improved, but device complexity increases
Solution Approach 1:
The patent extracts the complex computational processing function from the physical detector system and implements it as a separate artificial neural network. This allows the detector hardware to remain relatively simple while the ANN handles the computationally intensive real-time analysis, separating the complexity into a dedicated processing module.
Solution Approach 2:
The artificial neural network is pre-trained with simulation data before actual operation. This preliminary training phase allows the system to learn optimal processing patterns offline, so that during real-time operation, the system can quickly process detector signals without requiring complex real-time computational adjustments, thereby reducing operational device complexity.
3Device complexity
If data is processed remotely in a cloud-based computing system, then device complexity is reduced, but loss of time increases due to data transfer requirements
Solution Approach 1:
The patent introduces an artificial neural network as an intermediary processing layer between the detector and remote cloud systems. The ANN performs preliminary real-time processing locally, extracting essential radiation event information (position, energy, interaction type) before any remote data transfer, thereby minimizing the data volume and time required for cloud-based further analysis.
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 provides substantially real-time event characterization, improving accuracy and speed by monitoring radiation information event-by-event, compensating for signal fluctuations, and extracting interaction types, positions, and energies with high precision.
Implementation Method 1
a converter comprising a plurality of physically spaced semiconductor sensors configured to convert incident X-ray and/or gamma-ray photons into electron-hole pairs
Implementation Method 2
an electric field generator configured to apply an electric field to the plurality of physically spaced semiconductor sensors, thereby creating signals representative of a movement of charge carriers
Data Source
AI summary
A semiconductor radiation sensor device for characterizing X-ray and/or gamma-ray radiation has: a converter with physically spaced semiconductor sensors to convert incident X-ray and/or gamma-ray photons into electron-hole pairs; an electric field generator to apply an electric field to the sensors, thereby creating signals representative of a movement of charge carriers in the sensors; a readout circuitry to read out the signals; and a processing unit connected to the readout circuitry. The processing unit estimates an interaction time and a three-dimensional interaction position of an event in the converter by processing the sensors signals. The processing unit has an artificial neural network trained to generate the estimated interaction time and the three-dimensional interaction position of the event based on a model of physical and geometrical sensors properties and based on simulated time-varying sensors charges. A method of training an artificial neural network to generate a three-dimensional characterization is also provided.