Ultrasonic Imaging AI for Compound and Harmonic Images at High Frame Rates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Ultrasonic diagnostic apparatuses face a decline in frame rate and image quality, such as resolution and contrast, due to the need for multiple transmissions/receptions for compound processing and tissue harmonic imaging, especially when the object or probe moves during these processes.
Innovation Solution
An ultrasonic diagnostic apparatus that utilizes machine learning to generate compound processing-equivalent images and pseudo-tissue harmonic imaging using learned models, allowing for image quality improvement without the need for multiple scans, thereby maintaining or enhancing frame rate.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If compound processing is performed by transmitting/receiving ultrasonic pulses multiple times on the same location, then image quality (resolution and contrast) is improved, but frame rate declines
Solution Approach 1:
The patent uses a neural network to create a learned model that copies the characteristics of compound processed images from training data. During operation, the system transmits ultrasonic pulses only once and uses the learned model to generate estimated compound images, effectively copying the quality benefits of multi-scan processing without performing the actual multiple scans.
Solution Approach 2:
The patent replaces the mechanical process of performing multiple physical transmissions and receptions with a computational model. Instead of physically scanning the same location multiple times to accumulate data, the system uses a neural network model trained on such data to synthetically generate the improved image, substituting physical repetition with computational estimation.
2Manufacturing precision
If tissue harmonic imaging is performed by transmitting/receiving multiple waveforms, then resolution and penetration are improved, but frame rate and image quality (when object moves) decline
Solution Approach 1:
The patent trains a neural network on pairs of fundamental wave images and their corresponding harmonic images. During operation, the system transmits only the fundamental wave and uses the learned model to copy-generate the harmonic image, eliminating the need for actual harmonic transmission and reception multiple times.
Solution Approach 2:
The patent replaces the mechanical process of transmitting inverted waveforms and receiving harmonic signals with a computational neural network that estimates harmonic images from fundamental wave data. This substitution maintains resolution benefits while eliminating the time penalty of multiple transmissions.
3Manufacturing precision
If multiple transmissions/receptions are performed for compound processing, then image quality is improved, but the process becomes complex and time-consuming
Solution Approach 1:
The patent uses a neural network model that has learned the transformation from single-scan images to compound images during training. During operation, the system simply needs to input the single-scan image data and the neural network outputs the estimated compound image, copying the complex processing results without replicating the complex acquisition process.
Solution Approach 2:
The patent replaces the complex mechanical system of coordinating multiple transmissions, receptions, and signal combinations with a single neural network inference operation. The neural network encapsulates the complexity of multi-scan processing in its weights and architecture, transforming it from a runtime computational burden to a pre-learned pattern recognition task.
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 apparatus achieves favorable image quality with reduced frame rate decline by generating compound and harmonic images through estimation, improving resolution and contrast while maintaining or increasing the frame rate.
Implementation Method 1
the ultrasonic probe 102 transmits ultrasonic waves from the plurality of transducers 101 to the object 100 and receives reflected ultrasonic waves from the object 100
Implementation Method 2
receives reflected ultrasonic waves having been reflected inside the object 100
Implementation Method 3
Beamforming of a transmit beam is performed by inputting a voltage waveform provided with a time delay relative to a plurality of conversion elements and causing ultrasonic waves to converge inside a living organism
Implementation Method 4
In the pulse inversion method, by adding up received signals obtained by transmitting/receiving a first transmission waveform and a second transmission waveform created by inverting a phase of the first transmission waveform, a fundamental wave component is canceled and a harmonic component is enhanced
Implementation Method 5
The tissue harmonic imaging involves extracting a harmonic component caused by unique nonlinearity of an object (body tissue) from a received signal
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An ultrasonic diagnostic apparatus, comprising: an ultrasonic probe which scans an observation region in an object with an ultrasonic wave; and an estimated image generating means adapted, by using a model having been machine-learned using learning data including first data based on a first received signal that is obtained by first transmission/reception of an ultrasonic wave and second data based on a second received signal that is obtained by second transmission/reception that represents a larger number of transmissions/receptions than the first transmission/reception of the ultrasonic wave, to generate an estimated image equivalent to image data obtained by the second transmission/reception from third data based on a third received signal that is obtained by transmission/reception equivalent to the first transmission/reception of the ultrasonic wave by the ultrasonic probe.