Design and Progession Analysis of Neuro Rehabilitation System using Machine Learning

Project Title

Design and Progession Analysis of Neuro Rehabilitation System using Machine Learning

Project Area of Specialization NeuroTechProject Summary

As Neuro rehabilitation is a complex medical process designed to help to treat patients with neurological diseases, therefore the aim of the project is to reduce the complexity of system. Rehabilitation aims to increase function, reduce debilitating symptoms, and improve a patient’s quality of life. The incidence of many neurologic diseases is rising partly. As a result, more survivors are emerging with most exhibiting life altering impairments that require neurorehabilitation. Moreover, physiotherapy treatment is expensive plus traveling and frequent visits to the physiotherapy centers and hospitals are not feasible for everyone and time consuming too. Therefore, our aim is to design the Neuro Rehabilitation System which is home convenient, less complex, less costly and less time consuming.  

Stroke is the second leading cause of death and third most common contributor to disability and a costly disease. This is the reason we have chosen “Stroke Rehabilitation” as our probe domain.

Secondly, in this revolutionary era of technology, most of the doctors are still relying on Manual Muscle Testing (MMT) or manual instruments like goniometer. Although digital goniometers are there but they are quite expensive, are not widely used in hospitals in Pakistan and require expertise. Therefore, we propose an automated Neuro Rehabilitation system that utilizes two sensors namely Electromyography (EMG) and Inertial Measurement Unit (IMU) to collect the data of patients regarding the three important parameters.

• Range of Motion (ROM)
• Smoothness
• Muscle contraction

A suitable set of exercises in virtual environment will be performed by patients and our designed machine learning model will assess patient’s task-oriented rehabilitation exercises to conclude how each set of exercise could provide improvement in the performance of the patient.

As Machine learning models can easily utilize data of patients through sensors so instead of manually reviewing abundant features, machine intelligence can automatically identify salient features to provide useful insights on patient’s performance to validate the prediction of outcomes. A newly developed ML model for Embedded systems will be deployed on Microcontroller using TinyML to build our Neuro Rehabilitation System.

Project Objectives

“To design an Automated Neuro rehabilitation system with the aid of sensors and (ML) machine learning. The system which will be efficient, cheap, less complex and home convenient.”

1. To collect the data from case study of stroke’s patient through the sensors while performing set of exercises in virtual environment to make the environment and exercises easier and more fun. As a result, it will enhance the Neuro Rehabilitation System to an automized one instead of manual one depending on clinical expertise. 
2. To design the machine learning model based on Neural networks by deploying advanced and compact ML models using TinyML on embedded microcontroller. This is to derive the predicted rehabilitation outcomes i.e to get the prediction of end results that how each set of task-oriented exercise improved patients’ functionality.
3. To examine the evaluated results and make a clear comparison to the data evaluated from the manual systems through MMT and goniometer and provide clear insight into the clinical based Neuro Rehabilitation and our designed automated Neuro Rehabilitation. Along with the validation from the doctor.

Project Implementation Method

1. Our proposed Neuro Rehabilitation System will first start with data analysis by collecting data through two basic sensors EMG and IMU of case study of three stroke patients available at the local hospital (we are collaborating with Bhittai Institute of Physical Therapy and Rehabilitation sciences, Canal Road, Mirpurkhas, Sindh (Pakistan))
2. Suitable number of set of exercises will be performed on three patients of stroke.
3. Analysis of the following parameters will be done as feature extraction for ML model.

• ROM
• Smoothness
• Muscle contraction 

                                                                                                                         figure 1. set of exercises

 images/Design and Progession Analysis of Neuro Rehabilitation System using Machine Learning _1639949466.png

1. The designed ML model based on Neural network will do the analysis of particular set of exercises. Consequently, we will predict outcomes through ML that which set of exercise enhances the movement in case study of stroke’s patient with the help of data collected through the sensors. This will be the assessment of task-oriented exercises performed in home environment and will conclude the improvement in patient without the presence of doctor.
2. This will be done by using tinyML by deploying ML model on microcontroller using specific library CMISNN. Specially build for deploying NN models on embedded systems/ MCUs.
3. The data of the patient will be collected through conventional clinical method as well  depending on manual systems like goniometer to evaluate the results through standard tests.
4.  A clear insight for the comparison between our automated system to the    conventional clinical system will be provided. Our evaluated results will show how the system can perform effectively without the need of doctors, is less costly, home convenient, less complex, easily access the improvement in functionality through ML while being at home in comparison to the clinical one.                                                                                    figure 2. Steps of proposed methodologyimages/Design and Progession Analysis of Neuro Rehabilitation System using Machine Learning _1639949468.png

Benefits of the Project

• The system would be less complex in comparison to the clinical process.
• No need of clinical expertise to be physically present.
• Home convenient, time saving, less costly.
• The assessment of improvement of exercises on the patients will be concluded using machine Learning model which will be efficient and accurate instead of relying on manual systems to analyze the improvements.
• Implementing the new tinyml technology by deploying ML models on microcontroller which has significant benefits over the use of GPU, Cloud etc. in the field of health care.

Technical Details of Final Deliverable

This project comprises of tensor Flow lite for micro software and hardware, i.e sensors and Arduino Nano 33 BLE Sense /STM32F4 nucleo microcontroller, for deploying the machine learning model. Sensors are used for recognizing the parameters i.e ROM, smoothness and muscle contraction. Machine learning model will be designed on tensor flow lite for micro and deploy on microcontroller using tiny ml. Deployment of machine learning model on microcontroller has its own advantages over the Cloud, GPU like less power consumption, smaller in size, less costly. The end delivery will be closed loop hardware and software integrated system working together in real-time concluding which particular set of exercises would enhance the performance of particular stroke’s patient.

                      

                     Figure 3. Hardware and Software tools

images/Design and Progession Analysis of Neuro Rehabilitation System using Machine Learning _1639949470.png

Final Deliverable of the Project HW/SW integrated systemCore Industry HealthOther Industries Medical , Health Core Technology NeuroTechOther Technologies Artificial Intelligence(AI), Wearables and ImplantablesSustainable Development Goals Good Health and Well-Being for PeopleRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	72783
EMG Sensor	Equipment	2	3500	7000
IMU Sensor	Equipment	2	2000	4000
Arduino Nano 33 BLE Sense /STM32F4 nucleo Microcontroller	Equipment	1	22000	22000
VR SHINECON 3D VR Glasses Virtual Reality Headset Monochrome	Equipment	1	1338	1338
Monochrome o.96”OLED	Equipment	4	400	1600
PCB designing	Equipment	4	5000	20000
Plastic goniometer (for comparative testing)	Equipment	1	845	845
Consultation fee of physiotherapists and orthopedics (for testing)	Miscellaneous	1	5000	5000
TinyML Machine Learning with TensorFlow Lite on Arduino and Ultra-Low-	Equipment	1	6000	6000
Travelling, printing etc	Miscellaneous	1	5000	5000