Inverse neural network for particle detection in a solid-state-devices
AssigneeSiemens Medical Solutions USA, Inc.
InventorsSrutarshi Banerjee, Miesher Rodrigues
Life of this patent
Estimating the term…
Abstract
For training to and/or estimating location, energy level, and/or time of occurrence of incident radiation on a solid-state detector, a machine-learned model, such as a neural network, performs the inverse problem. An estimate of the location, energy level, and/or time is output by the machine-learned model in response to input of the detected signal (e.g., voltage over time). The estimate may account for material property variation of the solid-state detector in a rapid and easily calculated way, and with a minimal amount of data.
Claims17
A method of detecting incident radiation with a solid-state detector, the method comprising:
- measuring a first signal by the solid-state detector, the first signal generated in response to the incident radiation being received by the solid-state detector;
- inputting the measured first signal into a discriminator network configured to output a position and an energy level when input a measured signal from the solid-state detector, wherein the discriminator network is trained using a generative adversarial process as part of a generative adversarial network comprising the discriminator network and a generator network, wherein the generator network is a forward model that uses a position of an interaction and a number of charges to generate signal data, wherein the discriminator network is adversarially trained by inputting either the generated signal data or measured signal data and generating an estimated position and energy level;
- determining, by the discriminator network, a first position on the solid-state detector and a first energy level of the incident radiation on the solid-state detector from the measured first signal;
- and
- generating an image from the first position and the first energy level.
A method for machine training of inverse detection by a solid-state detector, the method comprising:
- receiving, by a processor, first samples of signals measured by the solid-state detector in response to input energies at input locations;
- receiving, by the processor, second samples of signals created by a machine-learned generator network in response to the input energies and the input locations, wherein the machine-learned generator network uses a forward model that uses the input energies and the input locations to generate sample signal data;
- machine training a discriminator network configured to output a position and an energy level when input a measured signal from the solid-state detector, wherein the discriminator network is trained using a generative adversarial process as part of a generative adversarial network comprising the discriminator network and the machine-learned generator network, wherein the discriminator network is adversarially trained by inputting either the first samples of signals or second samples of signals and generating an estimated position and energy level;
- and
- storing the machine-trained discriminator network.
A system for detection of incident radiation, the system comprising:
- a solid-state detector configured to generate signal data in response to receipt of the incident radiation;
- and
- a processor configured to determine a location of the incident radiation on the solid-state detector, the location determined by a machine-learned model in response to input of the signal data to the machine-learned model, wherein the machine-learned model comprises a discriminator network of a generative adversarial network, the generative adversarial network comprising a generator and the discriminator, wherein the discriminator network is trained using a generative adversarial process as part of the generative adversarial network, wherein the generator network is a forward model that uses a position of an interaction and a number of charges to generate signal data, wherein the discriminator network is adversarially trained by inputting either actual signal data or generated signal data from the generator network and generating an estimated location and energy level.
Dependencies are inferred from the claim text.