Ultrasound Data Generation Using Trained Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ultrasonic diagnostic apparatuses face challenges in achieving high-quality images with reduced sidelobe artifacts while maintaining a high frame rate, due to the need for multiple ultrasonic transmissions and the resulting limitations in bandwidth and axial resolution.
Innovation Solution
The ultrasonic diagnostic apparatus employs a data generation function using a trained model to generate output data based on a non-linear signal from input data acquired through a single ultrasonic transmission of a fundamental wave signal, thereby improving image quality without the need for multiple transmissions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If tissue harmonic imaging is performed using pulse inversion or amplitude modulation, then image quality is improved with reduced sidelobe artifacts, but the number of transmissions increases which lowers the frame rate
Solution Approach 1:
The patent creates a virtual copy of the harmonic imaging effect through signal processing. By training a neural network model on pairs of fundamental wave signals and corresponding harmonic signals, the system can generate harmonic-like output data from a single fundamental wave transmission, copying the effect of multiple transmissions without actually performing multiple transmissions.
Solution Approach 2:
The patent replaces the physical mechanism of multiple ultrasonic transmissions with a computational mechanism. Instead of using pulse inversion or amplitude modulation techniques that require multiple transmissions, the system uses a trained neural network model to process the received signal and generate harmonic imaging data through software-based signal transformation.
2Productivity
If filtering is used for tissue harmonic imaging, then the process can be completed in one transmission, but the bandwidth becomes too narrow which lowers the axial resolution
Solution Approach 1:
The patent changes the frequency parameters of the received signal through neural network processing. The model learns to transform the frequency content of the fundamental wave signal to generate harmonic frequency components, effectively expanding the usable bandwidth while maintaining high axial resolution through the computational generation of harmonic frequencies.
3Reliability
If multiple ultrasonic transmissions are conducted to acquire add signal, then signal-to-noise ratio is improved, but the number of transmissions increases which reduces the frame rate
Solution Approach 1:
The patent creates a virtual copy of the signal averaging effect. Instead of physically transmitting multiple ultrasonic waves and averaging the received signals, the neural network model processes a single received signal to generate output that has the statistical properties of averaged signals, copying the noise-reduction effect without requiring multiple transmissions.
Data Source
AI summary
According to one embodiment, an apparatus includes processing circuitry. The processing circuitry acquires output data from a trained model by entering examination data acquired at an examination, the examination data corresponding to first data, the output data corresponding to second data, into the trained model configured to, based on the first data acquired through transmission of an ultrasound wave for a first number of times, output the second data acquired through transmission of an ultrasound wave for a second number of times that is greater than the first number of times.