Assessing diseases by analyzing gait measurements
Inventors
MOTAMEDI, Gholam • Frieder, Ophir • Flagg, Cristopher • Wu, Jian-Young
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12690782-B2
Publication Date
2026-07-28
Expiration Date
Abstract
A gait analysis system, which includes a neural network with a recurrent neural network layer and a fully connected layer, that receives sensor data indicative of an individual's gait and outputs an assessment regarding the individual's health. The neural network is trained using training data indicative of abnormal gaits and normal gaits. To analyze the training data and the sensor data, the recurrent neural network layer parses each piece of data into a series of windows and analyzes each window in series to generate a context vector characterizing each window and the previously analyzed windows. The fully connected layer, having been trained to differentiate between normal gaits and abnormal gaits based on context vectors characterizing the training data, is used to generate a final assessment characterizing the user gate as normal or abnormal using one or more of the context vectors characterizing the sensor data.
Core Innovation
The invention stores gait data that includes training data indicative of abnormal gaits and normal gaits, and sensor data indicative of a user gait. It trains a neural network that includes a recurrent neural network layer and a fully connected layer by parsing the training data into a series of training data windows and analyzing each training data window in series to generate a context vector characterizing each training data window and previously analyzed training data windows. The fully connected layer is trained to distinguish between normal gaits and abnormal gaits based on one or more of the context vectors.
For assessment, the invention parses the sensor data into a series of sensor data windows and analyzes each sensor data window in series using the recurrent neural network layer to generate a context vector characterizing each sensor data window and previously analyzed sensor data windows. Generating the context vector includes generating an interim assessment of the normality of the user gait. The invention then generates a final assessment characterizing the user gait as normal or abnormal using the fully connected layer trained using the training data, based on one or more context vectors characterizing the sensor data.
In the dependent variations described with the same overall approach, the neural network can further include an attention layer that assigns relative importance to training data windows and/or sensor data windows. The recurrent neural network can be bidirectional to process sensor data windows to generate multiple context vectors that characterize preceding and subsequent windows.
Claims Coverage
The independent claims are clm-00001 and clm-00011. Across these claims, the inventive features focus on training a recurrent-neural-network-based system over windowed gait data to generate context vectors and using those context vectors in a fully connected layer to output an assessment of normality.
Windowed sequence training with recurrent context vectors and fully connected normal/abnormal distinction
The method trains a neural network including a recurrent neural network layer and a fully connected layer by parsing training data into a series of training data windows, analyzing each training data window in series via the recurrent neural network layer to generate a context vector characterizing each training data window and previously analyzed training data windows, and training the fully connected layer to distinguish between normal gaits and abnormal gaits based on one or more context vectors.
Interim and final normality assessment from sensor-window context vectors
The method parses sensor data into a series of sensor data windows, analyzes each sensor data window in series via the recurrent neural network layer to generate a context vector characterizing each sensor data window and previously analyzed sensor data windows, wherein generating the context vector comprises generating an interim assessment of the normality of the user gait, and generates a final assessment characterizing the user gait as normal or abnormal by the fully connected layer using one or more context vectors characterizing the sensor data.
Neural network system with recurrent context-vector generation and normal/abnormal output
The system includes non-transitory computer readable storage media storing training data indicative of abnormal gaits and normal gaits, and a hardware computer processor that stores a neural network with a recurrent neural network layer and a fully connected layer by parsing training data into a series of training data windows, analyzing each training data window in series via the recurrent neural network layer to generate a context vector characterizing each training data window and previously analyzed training data windows, and training the fully connected layer to distinguish between normal gaits and abnormal gaits based on one or more context vectors.
System receives sensor data and generates interim and final normal/abnormal assessments
The system receives sensor data indicative of a user gait, parses the sensor data into a series of sensor data windows, analyzes each sensor data window in series via the recurrent neural network layer to generate a context vector characterizing each sensor data window and previously analyzed sensor data windows, wherein generating the context vector comprises generating an interim assessment of the normality of the user gait, and generates a final assessment characterizing the user gait as normal or abnormal by the fully connected layer trained using the training data, using one or more context vectors characterizing the sensor data.
Across the independent claims, the core claim structure is the same: parse training data and sensor data into windows, use a recurrent neural network layer to generate context vectors that characterize each window, and use a fully connected layer to generate a final assessment of normal versus abnormal gait. Dependent claims indicate extensions to attention-weighted window importance, bidirectional recurrent processing, and broader assessments such as disease, disease stage, and injury timing based on context vectors.
Stated Advantages
Documented Applications
No documented applications found
Interested in licensing this patent?