Ultrasound system with target and medical instrument awareness
Blood vessel recognition and needle guidance systems, components, and methods thereof. A console can be configured to initiate a target recognition process for recognizing an anatomical target, such as a blood vessel, of a patient and a needle guidance process for guiding insertion of a needle into the anatomical target using ultrasound-imaging data received by the console. The system can perform target identification based on machine learning models which can be trained to recognize targets based on the ultrasound image. The ultrasound probe can be configured to provide to the console electrical signals corresponding to the ultrasound-imaging data. The ultrasound probe can include an array of transducers and, optionally, an array of magnetic sensors respectively configured to convert reflected ultrasound signals from the patient and magnetic signals from the needle, when magnetized, into the electrical signals.
1 . A medical method, comprising:
receiving user input indicating a desired target into a console including a processor and non-transitory, computer-readable medium having logic stored thereon;
initiating in the console:
a target recognition process to identify a blood vessel of a patient corresponding to the desired target by applying a machine learning model to features of candidate targets in ultrasound-imaging data, wherein the blood vessel identified as the desired target is associated with a highest confidence score determined by the machine learning model; and
a needle guidance process for guiding insertion of a needle into the blood vessel identified as the desired target by the machine learning model, wherein the machine learning model determines a proximity of a trajectory of the needle to the blood vessel identified as the desired target based on magnetic or optical information about the needle;
loading the ultrasound-imaging data in the console, the ultrasound-imaging data corresponding to electrical signals received from an ultrasound probe, wherein the electrical signals are produced from reflected ultrasound signals from the patient;
processing the ultrasound-imaging data with the console according to the target recognition process and the needle guidance process; and
inserting the needle into the blood vessel identified as the desired target.
2 . The method according to claim 1 , wherein the machine learning model includes a supervised learning model trained based on previously-identified training targets.
3 . The method according to claim 2 , wherein the supervised learning model is trained by iteratively minimizing an error, for classifications predicted by a candidate model compared to actual classifications of the previously-identified training targets.
4 . The method according to claim 1 , wherein each of the candidate targets in the ultrasound-imaging data is associated with a confidence score determined by the machine learning model.
5 . The method according to claim 1 , wherein recognizing the desired target using the machine learning model further comprises applying, by the console, one or more of the following supervised learning methods:
logistic regression;
other linear classifiers;
support vector machines;
quadratic classifiers;
kernel estimation;
decision trees;
neural networks; or
learning vector quantization.
6 . The method according to claim 5 , wherein the machine learning model includes the logistic regression, and wherein a confidence score of a first candidate target is determined as a weighted sum of the features of the candidate targets.
7 . The method according to claim 5 , wherein the machine learning model includes a first neural network, and wherein a confidence score of a first candidate target is determined based on one or more nonlinear activation functions of the features of the candidate targets.
8 . The method according to claim 1 , wherein the blood vessel comprises a superior vena cava.
9 . The method according to claim 1 , further comprising providing, by the ultrasound probe, a first ultrasound image prior to the insertion of the needle, wherein the machine learning model is based in part on the trajectory of the needle.
10 . The method according to claim 1 , wherein the features of the candidate targets include one or more of the following features:
a shape of a candidate blood vessel of the candidate targets in the ultrasound-imaging data,
a size of the candidate blood vessel,
a number of branches of the candidate blood vessel, or
a complexity of branches of the candidate blood vessel.
11 . The method according to claim 1 , wherein recognizing the blood vessel identified as the desired target comprises providing identifying information characterizing at least a first candidate target to a user, wherein the identifying information comprises a predicted classification of at least the first candidate target and a confidence level associated with the predicted classification.
12 . The method according to claim 1 , wherein guiding the insertion of the needle further comprises graphically guiding, by a display screen of the console, the insertion of the needle into the blood vessel identified as the desired target.
13 . The method according to claim 1 , wherein the machine learning model determines the proximity of the trajectory of the needle to the blood vessel identified as the desired target based on magnetic information about the needle, wherein the magnetic information about the needle is provided by magnetic sensors in the ultrasound probe.
14 . The method according to claim 1 , further comprising magnetizing the needle with a needle magnetizer incorporated into the console when the needle is inserted into the needle magnetizer.
15 . The method according to claim 14 , wherein the needle guidance process for guiding insertion of the magnetized needle into the blood vessel identified as the desired target uses a combination of the ultrasound-imaging data and magnetic-field data received by the console.
16 . The method according to claim 15 , wherein the ultrasound probe is configured to provide to the console the magnetic-field data, the ultrasound probe further including an array of magnetic sensors configured to convert magnetic signals from the magnetized needle into a magnetic-field portion of the electrical signals.
17 . The method according to claim 1 , further comprising implanting a catheter into the blood vessel identified as the desired target.
18 . The method according to claim 17 , further comprising:
loading, in a memory, electrocardiography information corresponding to electrical signals received from one or more electrocardiography probes; and
initiating, by the processor of the console, a catheter tip placement confirmation process based on the electrocardiography information.
19 . The method according to claim 17 , further comprising tracking the catheter using an optical tip-tracking system, wherein the optical tip-tracking system comprises a stylet and a light detector coupled to the console.
20 . The method according to claim 19 , wherein the stylet of the optical tip-tracking system comprises an optical fiber.