A Deep Learning-based Comprehensive Robotic System for Lower Limb Rehabilitation

Abstract

ABSTRACT A Deep Learning-based Comprehensive Robotic System for Lower Limb Rehabilitation In this invention, we introduce a deep learning-based robotic system designed to assist with lower limb rehabilitation, specifically targeting individuals experiencing knee pain. The system leverages a novel combination of EEG and EMG signals, along with knee bending angle data, to determine the pain level of patients using an advanced CNN-TLSTM network with an integrated attention mechanism. This deep learning model effectively classifies knee pain into five categories—no pain, low pain, medium pain, moderate pain, and high pain—achieving an overall classification accuracy of 95.88%. The innovative application of deep learning enables precise and efficient pain level assessment, facilitating tailored rehabilitation treatments that minimize the need for constant human supervision. The proposed robotic system features an adaptive exercise module that assists patients with three essential knee rehabilitation exercises: sitting knee bending, straight leg rise, and active knee bending. The system customizes exercise parameters, such as intensity and duration, in real-time based on the patient’s assessed pain level, ensuring that each session is optimized for individual needs and comfort. This personalized approach reduces the risk of overexertion while enhancing recovery outcomes. By automating rehabilitation exercises, the system provides consistent and accurate guidance, creating a safe and efficient rehabilitation environment for patients without the constant presence of a physiotherapist. Our system also integrates a feedback mechanism that records patient responses and adapts future sessions for improved effectiveness. This adaptability, along with the system's capacity to incorporate additional sensors like Kinect and ECG for comprehensive monitoring, paves the way for enhanced rehabilitation protocols. The intelligent framework offers a robust and scalable solution for knee pain management and recovery, allowing patients to undergo effective lower limb rehabilitation independently, and promoting more accessible and efficient treatment options across various healthcare settings.

Specification

Description:A Deep Learning-based Comprehensive Robotic System for Lower Limb Rehabilitation

This invention relates to the field of medical rehabilitation, specifically to a deep learning-based robotic system that provides lower limb rehabilitation for patients experiencing knee pain. The system employs an advanced artificial intelligence model to assess knee pain levels and administer targeted rehabilitation exercises without the need for a physiotherapist.

BACKGROUND
[0001] Background description includes information that may be useful in understanding the present invention. It is not an admission that any of the information provided herein is prior art or relevant to the presently claimed invention, or that any publication specifically or implicitly referenced is prior art.
[0002] Knee pain is one of the most common health issues affecting a large segment of the global population. It is often caused by a variety of conditions such as arthritis, ligament injuries, or repetitive stress, and can significantly affect an individual's mobility and quality of life. Effective rehabilitation plays a critical role in alleviating knee pain and restoring proper function. Traditional rehabilitation methods rely heavily on physical therapists to guide patients through exercise regimens, which can be time-consuming and require frequent visits to healthcare centers. This has led to a demand for more automated solutions that can offer patients continuous care in the comfort of their homes.
[0003] Recent advancements in technology have shown promise in improving rehabilitation outcomes through the use of robotics and artificial intelligence (AI). The application of AI, particularly deep learning models, has enabled the development of systems capable of assessing and monitoring patient conditions with high accuracy. These AI-powered systems have the potential to personalize rehabilitation exercises based on real-time data from patients, providing individualized treatment plans that are adjusted according to the patient's pain levels and recovery progress. Despite these advances, there remains a need for an integrated system that not only detects knee pain but also provides a comprehensive robotic solution for rehabilitation.
[0004] The challenge of accurately detecting and assessing knee pain is significant, as pain perception varies from person to person and is influenced by multiple physiological factors. Traditional methods of pain assessment, such as self-reports or manual evaluation by clinicians, are subjective and may not capture the true extent of the pain or its fluctuations over time. To address this issue, researchers have begun exploring the use of multi-modal data, including electroencephalogram (EEG) signals from the brain, electromyography (EMG) signals from the muscles, and knee joint angle data. These signals, when processed through deep learning models, can provide more accurate, objective, and real-time assessments of a patient's knee pain level.
[0005] In this context, the integration of a CNN-TLSTM (Convolution Neural Network - Tanh Long Short-Term Memory) network with an attention mechanism represents a breakthrough in pain detection. This model is capable of handling complex data from multiple sources, including EEG, EMG, and knee angle data, to assess pain levels in a reliable and scalable manner. By focusing on the most relevant features of the data through an attention mechanism, the model enhances classification accuracy and ensures that pain levels are categorized into clear categories, ranging from no pain to high pain. Such advancements in AI are key to developing more effective rehabilitation tools that can autonomously adapt to a patient's needs.
[0006] The design of a robotic rehabilitation system that adjusts its parameters based on the detected pain level marks another significant advancement. Rehabilitation devices that can automatically tailor exercise intensity, duration, and frequency based on real-time assessments of pain can reduce the risk of overexertion or injury. Furthermore, such systems eliminate the need for constant supervision by a physiotherapist, enabling patients to undergo rehabilitation independently. This opens the door for a more accessible and scalable approach to knee pain rehabilitation, which could have a profound impact on the way knee pain is managed and treated across various healthcare settings. As the technology continues to evolve, it holds promise for expanding to other forms of musculoskeletal rehabilitation.
[0007] All publications herein are incorporated by reference to the same extent as if each individual publication or patent application were specifically and individually indicated to be incorporated by reference. Where a definition or use of a term in an incorporated reference is inconsistent or contrary to the definition of that term provided herein, the definition of that term provided herein applies and the definition of that term in the reference does not apply.
[0008] As used in the description herein and throughout the claims that follow, the meaning of "a," "an," and "the" includes plural reference unless the context clearly dictates otherwise. Also, as used in the description herein, the meaning of "in" includes "in" and "on" unless the context clearly dictates otherwise.
OBJECTS OF THE INVENTION
[0009] It is an object of the present disclosure to provide a deep learning-based robotic system for lower limb rehabilitation that accurately assesses knee pain levels using a CNN-TLSTM network with an attention mechanism, leveraging multi-modal data such as EEG, EMG, and knee bending angle to classify pain levels into distinct categories: no pain, low pain, medium pain, moderate pain, and high pain.
[0010] It is an object of the present disclosure to design an adaptive robotic rehabilitation system that assists patients with three specific knee rehabilitation exercises-sitting knee bending, straight leg rise, and active knee bending-while dynamically adjusting exercise parameters such as intensity, duration, and frequency based on the patient's assessed pain level.
[0011] It is an object of the present disclosure to ensure the system's capability for personalized rehabilitation, providing patients with tailored exercise regimens that are continuously updated based on real-time data and individual progress, thus improving the efficiency and effectiveness of rehabilitation sessions.
[0012] It is an object of the present disclosure to develop a user-friendly system that automates knee rehabilitation without requiring the presence of a physiotherapist, allowing patients to perform exercises independently and safely while ensuring proper guidance and assistance through robotic intervention.
[0013] It is an object of the present disclosure to enable real-time patient monitoring and feedback integration, allowing the system to adapt future rehabilitation sessions based on patient responses and progress, ensuring a scalable and long-term solution for knee pain management and rehabilitation.
SUMMARY
[0001] The present invention presents deep learning-based comprehensive robotic system for lower limb rehabilitation.
[0002] This invention introduces a deep learning-based robotic system designed for lower limb rehabilitation, specifically targeting knee pain. The system uses a CNN-TLSTM (Convolution Neural Network - Tanh Long Short-Term Memory) network with an integrated attention mechanism to assess the pain level of the patient by analyzing multi-modal data such as EEG signals, EMG signals, and knee bending angle. The deep learning model classifies knee pain into five distinct categories: no pain, low pain, medium pain, moderate pain, and high pain, achieving a high classification accuracy of 95.88%. This classification is used to personalize and adapt rehabilitation exercises to each patient's pain level.
[0003] The robotic rehabilitation system assists patients in performing three key exercises-sitting knee bending, straight leg rise, and active knee bending-automatically adjusting the intensity, duration, and frequency of the exercises based on real-time pain assessments. This system allows patients to undergo rehabilitation without needing a physiotherapist, improving both the accessibility and efficiency of rehabilitation. With continuous monitoring and adaptive feedback, the system provides a scalable solution for knee pain rehabilitation, ensuring patients receive appropriate care while improving long-term recovery outcomes.
[0004] One should appreciate that although the present disclosure has been explained with respect to a defined set of functional modules, any other module or set of modules can be added/deleted/modified/combined and any such changes in architecture/construction of the proposed method are completely within the scope of the present disclosure. Each module can also be fragmented into one or more functional sub-modules, all of which also completely within the scope of the present disclosure.
[0005] Various objects, features, aspects and advantages of the inventive subject matter will become more apparent from the following detailed description of preferred embodiments, along with the accompanying drawing figures in which like numerals represent like components.
BRIEF DESCRIPTION OF THE DRAWINGS
[0014] The accompanying drawings are included to provide a further understanding of the present disclosure, and are incorporated in and constitute a part of this specification. The drawings illustrate exemplary embodiments of the present disclosure and, together with the description, serve to explain the analysis of the present disclosure.
[0015] Figure 1: A Deep Learning-based Comprehensive Robotic System for Lower Limb Rehabilitation.
DETAILED DESCRIPTION
[0016] In the following description, numerous specific details are set forth in order to provide a thorough understanding of embodiments of the present invention. It will be apparent to one skilled in the art that embodiments of the present invention may be practiced without some of these specific details.
[0017] If the specification states a component or feature "may", "can", "could", or "might" be included or have a characteristic, that particular component or feature is not required to be included or have the characteristic.
[0018] Exemplary embodiments will now be described more fully hereinafter with reference to the drawings, in which exemplary embodiments are shown. This disclosure, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. These embodiments are provided so that this disclosure will be thorough and complete and will fully convey the scope of the invention to those of ordinary skill in the art. Moreover, all statements herein reciting embodiments of the invention, as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Additionally, it is intended that such equivalents include both currently known equivalents as well as equivalents developed in the future (i.e., any elements developed that perform the same function, regardless of structure.
[0019] various terms as used herein are shown below. To the extent a term used in a claim is not defined below, it should be given the broadest definition persons in the pertinent art have given that term as reflected in printed publications and issued patents at the time of filing.
[0020] The robotic system uses a deep learning-based CNN-TLSTM model for real-time pain level assessment. The CNN component processes the incoming EEG and EMG signals alongside knee bending angle data to autonomously extract meaningful features. The TLSTM classifier, enhanced with an attention mechanism, interprets these features to classify knee pain into five categories. The integration of attention into the TLSTM allows the model to focus on critical signal segments, enhancing classification performance and yielding an overall improved accuracy.
[0021] The robotic system administers three main knee rehabilitation exercises: (i) sitting knee bending, (ii) straight leg rise, and (iii) active knee bending. Based on the classified pain level, the system automatically adjusts the exercise intensity, frequency, and duration to accommodate the patient's specific condition. By eliminating the need for a physiotherapist, the device offers a user-friendly interface and assists patients in conducting exercises independently.
[0022] A prototype of the rehabilitation device has been developed to support the three knee exercises. The hardware includes motorized supports and joint sensors that guide and monitor the patient's movements, ensuring proper posture and alignment. These elements work in unison with the pain detection model, enabling automated adjustments to exercise parameters in real-time.
[001] Data Collection and Input (100): The system collects real-time data from multiple sources, including EEG signals from the brain, EMG signals from the hamstring and quadriceps muscles, and knee bending angle measurements from sensors attached to the patient. This data serves as input for pain assessment and rehabilitation decision-making.
[002] Preprocessing and Feature Extraction (101): The raw data is preprocessed and features are extracted using a Convolutional Neural Network (CNN) model. The CNN is responsible for identifying important patterns in the EEG, EMG, and knee angle data, which are relevant to accurately assessing the pain level.
[003] Pain Level Classification (102): The extracted features are passed to a Tanh Long Short-Term Memory (TLSTM) network with an attention mechanism. The TLSTM model processes temporal data and classifies the pain level into one of five categories: no pain, low pain, medium pain, moderate pain, or high pain.
[004] Dynamic Exercise Adjustment: (103): Based on the classified pain level, the system adjusts the rehabilitation exercises-sitting knee bending, straight leg rise, and active knee bending-automatically. The exercise parameters such as intensity, duration, and frequency are customized to the patient's pain level to ensure safe and effective rehabilitation.
[005] Real-Time Monitoring and Feedback (104): During rehabilitation, the system continuously monitors the patient's response to the exercises and adjusts the parameters in real time. If there are any changes in the pain level or discomfort, the system adapts the exercises accordingly to maintain optimal rehabilitation.
[006] Patient Progress Tracking and Reporting (105): The system tracks the patient's progress over time and generates reports based on exercise performance and pain level fluctuations. This allows the patient and healthcare provider to monitor recovery progress, make informed decisions, and adjust treatment plans as necessary.
Figure 2 illustrates a process flow for knee pain classification using EEG and EMG signals. It begins with subject selection, followed by data acquisition that records EEG signals from specific electrodes, EMG signals from leg muscles, and knee bending angle. Pre-processing steps such as filtering and sampling prepare the data for dataset preparation. The dataset is then used to design and train a CNN-TLSTM model, whose performance is evaluated based on accuracy, recall, specificity, precision, and F1 score.
Figure 3 presents the architecture of a CNN-TLSTM model for pain level prediction. It begins with an input layer followed by two 1D convolution layers and a dropout layer, then passes through a 1D global average pooling layer and a flatten layer. The TLSTM layers with multiple units incorporate an attention mechanism, which combines hidden states to enhance feature selection. Finally, a fully connected layer, softmax layer, and output layer predict various pain levels such as no pain, low, medium, moderate, and high pain.
Figure 4 depicts a system for assisting patients in performing exercises based on their pain levels. The process begins with user inputs, including pain level, exercise name, and number of repetitions, which are sent to a control unit. The control unit, referencing an exercise database, directs the angular movement of actuators. These actuators control the movement of two metal links, enabling the patient to perform the prescribed exercises.
, Claims:I/We Claim
Claim 1: A deep learning-based robotic system for lower limb rehabilitation, comprising a pain detection module, a robotic exercise assistance module, a patient monitoring module, a rehabilitation adaptation module, and a user feedback module.
Claim 2: The system of claim 1, wherein the pain detection module utilizes a CNN-TLSTM network with an attention mechanism to analyze EEG and EMG signals, as well as knee bending angle data, to classify knee pain levels into five categories: no pain, low pain, medium pain, moderate pain, and high pain.
Claim 3: The system of claim 1, wherein the robotic exercise assistance module automatically adjusts and assists with three rehabilitation exercises-sitting knee bending, straight leg rise, and active knee bending-based on the patient's classified pain level.
Claim 4: The system of claim 1, wherein the patient monitoring module continuously captures EEG and EMG signals from targeted muscle areas, along with knee bending angles, to provide real-time data for assessing changes in the patient's pain levels.
Claim 5: The system of claim 1, wherein the rehabilitation adaptation module dynamically modifies exercise intensity, duration, and frequency according to the patient's pain level and recovery progress, ensuring customized and effective rehabilitation.
Claim 6: The system of claim 1, wherein the user feedback module collects data on patient responses and satisfaction, using this information to refine exercise parameters and improve system performance for future rehabilitation sessions.

Documents