Machine learning to extract quantitative biomarkers from ultrasound RF spectrums
The present disclosure provides for ultrasound systems and methods to pre-process ultrasound data to distinguish abnormal tissue from normal tissue. An exemplary method can include receiving a set of ultrasound data and partitioning the set into a set of windows. The method can then provide for processing the set of windows to determine a power spectrum for each window. The power spectrum for each window can be processed to determine a normalized power spectrum for each window. This normalized power spectrum can be processed for each window with a machine learning model. The method can then provide for displaying an image where each window of the set of windows is displayed using a unique identifier based on the output of the machine learning model.
1. An ultrasound system comprising:
a transducer configured to output a single static set of raw ultrasound radio frequency (RF) data forming a single frame of a tissue area;
a memory containing machine readable medium comprising machine executable code having stored thereon instructions;
a signal processing unit comprising one or more processors coupled to the memory, the one or more processors configured to execute the machine executable code to the cause the one or more processors to:
receive the single static set of raw ultrasound RF data;
process the single static set of raw RF ultrasound data with a machine learning model trained from a training data set having a set of B-mode ultrasound images derived from raw RF ultrasound data; and
output classification of the raw RF ultrasound data from an output of the machine learning model.
2. The ultrasound system of claim 1 , wherein the machine learning model is a k-means model or a model from a deep learning network comprising one or a combination to the algorithm types of CNN (Convolutional Neural Network), RBM (Restricted Boltzmann Machine), LSTM (Long Short Term Memory) or CapsNet (Capsule Networks).
3. The ultrasound system of claim 1 , wherein the training data set includes the B-mode ultrasound images labeled by a clinical standard.
4. The ultrasound system of claim 1 , wherein the machine learning model is an unsupervised model.
5. The ultrasound system of claim 1 , further comprising a display, the signal processing unit is operable to display an image on the display representing the set of ultrasound data using a unique identifier based on the output of the machine learning model.
6. The ultrasound system of claim 1 , wherein the classification is a clinical standard.
7. The ultrasound system of claim 1 , wherein the tissue area is from a liver of a patient.
8. The ultrasound system of claim 1 , wherein the signal processing unit is operable to:
partition the single static set of raw ultrasound RF data into a set of windows or bounding boxes; and
process the set of windows or bounding boxes to output a power spectrum for each window or bounding box of the set of windows of bounding boxes.
9. The ultrasound system of claim 8 , wherein the signal processing unit is operable to process the power spectrum to determine a normalized power spectrum.
10. The ultrasound system of claim 8 , wherein the power spectrum is taken using a continuous Fast Fourier Transform (FFT) or a discrete FFT.
11. The ultrasound system of claim 1 , wherein the machine learning model outputs a specific multi-parametric biomarker.
12. A method comprising:
receiving a single static set of raw ultrasound radio frequency (RF) data output from at least one ultrasound transducer, the single static set of raw ultrasound RF data forming a single frame representing a tissue area of a patient;
inputting the single static set of raw ultrasound RF data as an input into a machine learning model trained from a training data set having a set of B-mode ultrasound images derived from raw RF ultrasound data; and
outputting a classification of the raw RF ultrasound data from an output of the machine learning model.
13. The method of claim 12 , wherein the machine learning model is a k-means model or a model from a deep learning network comprising one or a combination to the algorithm types of CNN (Convolutional Neural Network), RBM (Restricted Boltzmann Machine), LSTM (Long Short Term Memory) or CapsNet (Capsule Networks).
14. The method of claim 12 , wherein the training data set includes the B-mode ultrasound images labeled by a clinical standard.
15. The method of claim 12 , wherein the classification is a clinical standard.
16. The method of claim 12 , further comprising:
partitioning the single static set of raw ultrasound RF data into a set of windows or bounding boxes; and
processing the set of windows or bounding boxes to output a power spectrum for each window or bounding box of the set of windows or bounding box.
17. The method of claim 16 , wherein the step of processing the set of windows comprises processing the set of windows to output a time frequency domain processing technique.
18. The method of claim 16 , wherein the method further comprises processing the power spectrum to determine a normalized power spectrum.
19. The method of claim 16 , wherein the power spectrum is taken using a continuous Fast Fourier Transform (FFT) or a discrete FFT.
20. A method comprising:
receiving static sets of raw ultrasound radio frequency (RF) data output from at least one ultrasound transducer forming a single frame representing a tissue area of a patient;
determining a B-mode image output from the static sets of raw ultrasound RF data;
creating a training data set for a machine learning model with the inputs of static sets of raw ultrasound RF data and the B-mode image outputs; and
training the machine learning model to process an input of a single static set of ultrasound RF data to output a classification of the tissue area from the machine learning model.