Technologies for monitoring and predicting impaction state of an orthopaedic surgical implement during an orthopaedic surgical procedure
Technologies for monitoring impaction and predicting impaction state during an orthopaedic surgical procedure include one or more impaction sensors that generate sensor data. The surgical procedure includes impaction of an orthopaedic implement such as a surgical instrument or a prosthetic component. An impaction analyzer generates an impaction state prediction with a machine learning model based on the sensor data. The impaction state prediction may include an unseated state, a seated state, and a fracture state. An impaction state user interface outputs the impaction state prediction. A model trainer may train the machine learning model with labeled sensor data.
1 . A method for predicting impaction state during an orthopaedic surgical procedure, the method comprising:
collecting sensor data during the orthopaedic surgical procedure from an impaction sensor, wherein the sensor data is indicative of impaction state of an orthopaedic implement relative to a patient's bone;
generating an impaction state prediction with a machine learning model based on the sensor data, wherein the impaction state prediction comprises an unseated state, a seated state, or a fracture state; and
outputting the impaction state prediction.
2 . The method of claim 1 , wherein collecting sensor data from the impaction sensor comprises:
collecting vibration data from a vibration sensor coupled to a surgical instrument;
collecting motion data from an inertial measurement unit coupled to the surgical instrument; and
collecting audio data from an external microphone.
3 . The method of claim 1 , wherein generating the impaction state prediction with the machine learning model based on the sensor data comprises:
pre-processing the sensor data to generate processed sensor data; and
inputting the processed sensor data to the machine learning model.
4 . The method of claim 1 , wherein generating the impaction state prediction with the machine learning model based on the sensor data comprises:
inputting the sensor data to a recurrent neural network to generate anomaly prediction data; and
inputting the anomaly prediction data to a classifier to generate the impaction state prediction.
5 . The method of claim 4 , wherein the recurrent neural network comprises a long short-term memory network, and wherein the classifier comprises a random forest predictive model.
6 . A method for training a machine learning model for impaction state prediction, the method comprising:
collecting sensor data from an impaction sensor, wherein the sensor data is indicative of impaction state of an orthopaedic implement relative to a bone or bone analog;
labelling the sensor data with an impaction state label to generate labeled sensor data, wherein the impaction state label comprises an unseated state, a seated state, or a fracture state; and
training a machine learning model to predict impaction state for input sensor data based on the labeled sensor data.
7 . The method of claim 6 , wherein training the machine learning model based on the labeled sensor data comprises:
pre-processing the labeled sensor data to generate processed sensor data; and
training the machine learning model based on the processed sensor data.
8 . The method of claim 7 , wherein pre-processing the labeled sensor data comprises:
transforming the labeled sensor data to a frequency domain to generate frequency domain sensor data; and
reducing the dimensionality of the frequency domain sensor data to generate the processed sensor data.
9 . The method of claim 8 , wherein reducing the dimensionality of the frequency domain sensor data comprises performing principal component analysis of the frequency domain sensor data.
10 . The method of claim 6 , wherein training the machine learning model based on the labeled sensor data comprises:
training a recurrent neural network with the labeled sensor data to identify anomalies in the labeled sensor data; and
training a classifier with the anomalies in the labeled sensor data to predict the impaction state.
11 . The method of claim 10 , wherein the recurrent neural network comprises a long short-term memory network.
12 . The method of claim 11 , wherein the classifier comprises a random forest predictive model.
13 . A method for training an impaction analyzer for an orthopaedic surgical procedure, the method comprising:
generating, by an impaction sensor, sensor data indicative of impaction state of an orthopaedic implement relative to a bone or bone analog;
collecting the sensor data from the impaction sensor;
labeling the sensor data with an impaction state label to generate labeled sensor data; and
training a machine learning model to predict impaction state for input sensor data based on the labeled sensor data.
14 . The method of claim 13 , wherein collecting the sensor data from the impaction sensor comprises collecting vibration data from a vibration sensor coupled to a surgical instrument.
15 . The method of claim 13 , wherein collecting the sensor data from the impaction sensor comprises to collecting motion data from an inertial measurement unit coupled to a surgical instrument.
16 . The method of claim 13 , wherein to collecting the sensor data from the impaction sensor comprises to collecting audio data from an external microphone.
17 . The method of claim 13 , wherein to collecting the sensor data from the impaction sensor comprises to collecting sensor data from a sensor coupled to a surgical instrument, wherein the sensor comprises a force sensing resistor, a load cell, or a displacement sensor.
18 . The method of claim 13 , wherein training the machine learning model based on the labeled sensor data comprises:
transforming the labeled sensor data to a frequency domain to generate frequency domain sensor data;
reducing dimensionality of the frequency domain sensor data to generate the processed sensor data; and
training the machine learning model based on the processed sensor data.
19 . The method of claim 18 , wherein reducing the dimensionality of the frequency domain sensor data comprises performing principal component analysis of the frequency domain sensor data.
20 . The method of claim 13 , wherein training the machine learning model to predict impaction state for input sensor data based on the labeled sensor data comprises:
training a recurrent neural network with the labeled sensor data to identify anomalies in the labeled sensor data; and
training a classifier with the anomalies in the labeled sensor data to predict the impaction state.