23 March 2025

Introduction

My colleague Lee Yeonjae and I had the opportunity to attend a cohesion lunch with venture capitalists, entrepreneurs, and mentors. While mingling, we met undergraduate students already launching their own startups. Inspired by their drive, we realized that nothing was stopping us from initiating something meaningful before graduation. With a shared passion for self-initiated projects, we brainstormed ideas that could make a significant impact.

Modern lifestyles often lead to prolonged poor posture, which can result in disc conditions like bulging or herniated discs. These issues can compress spinal nerves, causing pain and discomfort.

SENSE logo

SENSE is a medical device we’re developing to detect abnormal muscle activity using surface electromyography (sEMG) sensors. It identifies disc conditions by:

1. Collecting voltage signals from muscles
2. Amplifying and processing the signals
3. Converting signals into 2D images using transformation techniques (e.g., Continuous Wavelet Transform, Empirical Mode Decomposition)
4. Feeding the images into a trained convolutional neural network (CNN) model
5. Diagnosing and prognosticating disc herniation through binary classification

Basic CNN architecture

The CNN model is trained on data from patients and healthy subjects. Although SENSE is still in development, this blog post will document our progress, insights, challenges, and any other exciting updates.

Stacking sEMG signal based on gait cycle segmentation

We are also looking into integrating inertial motion unit (IMU) into the signal processing flow. One way is to integrate the IMU signal directly by transforming it into 2D image like representation, just like how sEMG data is processed. On the other hand, we are also exploring potential to utilize IMU to segment the data based on one gait cycle, which can be fed into the network.

Current Progress

Our idea began as a posture correction device, alerting users to avoid prolonged bad postures. However, after discussions and consultations, we pivoted multiple times, eventually arriving at SENSE. Throughout this process, we conducted extensive research, from acquiring background knowledge to reading up on the latest technologies. We sought advice from NTU professors, orthopedic clinicians, and business owners, building connections with experts willing to support our project, including:

• Dr. Yeo Si Yong: Assistant Professor, Lee Kong Chian School of Medicine
• Dr. Jiang Lei: Consultant Spine Surgeon, Singapore General Hospital
• Dr. Jack Lo Yu Tung: Neurosurgery Resident, National Neuroscience Institute (NNI) Singapore
• Dr. Lam Wee Leon: Senior Consultant and Hand ＆ Reconstructive Microsurgeon, Singapore General Hospital
• Dr. Chen Jing: Rehabilitation Medicine Consultant, Singapore General Hospital
• Dr. Tan Ia Choo Celia: Clinical Assistant Professor, SingHealth Duke-NUS Surgery Academic Clinical Programme

Hardware setting

Static Proof of Concept

Model Accuracy

Model Loss

Yeonjae and I dedicated ourselves to this project, achieving significant milestones:

• Development of hardware and software for proof of concept
• Successful static environment proof of concept by simulating abnormal muscle activities using electrical stimulations
• Establishing collaboration with doctors from Singapore General Hospital
• Successful classification model proof of concept using publicly available sEMG dataset
• Awarded multidisciplinary team fund of up to S$ $10,000 for the project
• Awarded academic clinical program pitch-for-funds of $100,000 for the project

You may find all our code and results here. For more details on our progress, refer to our meeting and progress records.

Recently, we have submitted an IRB application to SGH, which will start its review process by the end of March 2025. We also succesfully received the ACP pitch-for-funds, which will fund S$100,000 for our project. We have started planning on allocation of budget for things like equipment purchases, recruitment of research assistants, and facility lending.

Furthermore, we were introduced to professor Celia, who is part of the academic clinical programme. She had Cometa sEMG and and Xsens IMU, which she agreed to lend our team for experimentation purposes. We will be using this equipment to plan and execute mock tests to validate the integration of IMU to improving classification accuracies.

In an extension to our study, we were invited to take part in writing a review article on current trends of neuroprosthetics. Dr. Leon, who invited us to participate, suggested that since our project scope lies in the application of sEMG sensors, we might explore sEMG used in neuroprosthesis control. Title of the paper is Advances in Myoelectric Neuroprosthetics: A Narrative Review of Current Trends, Challenges, and Future Directions. Yeonjae and I are the first author of the paper, and the paper is current under review. It is expected to be published by the end of April 2025.

Ongoing Work

Our to-do list is constantly evolving. Current tasks include:

• Review IRB application for approval
• Designing and executing mock data collection with sEMG and IMU sensors
• Improving the model by using different architectures (RNN, transformers) and attempting to integrate IMU data into the signal processing flow

In the long term, we aim to:

• Collect data on disc patients and healthy subjects
• Develop affordable suit like sEMG and IMU sensor hardware that connect with a mobile app for disc condition diagnosis
• Pursue patent or other IP protection
• Productize the device
• Publish our findings

Challenges

The SENSE project has been fraught with challenges, but these obstacles have fueled our growth.

Our biggest challenge was the lack of expertise. As mechanical engineering students, we had limited knowledge in medical, computer science, and electrical fields. To bridge this gap, we invested considerable time in research, reading papers, and seeking guidance from experts. We reached out to professors and clinicians both locally and internationally, and even contacted authors of relevant papers. These efforts taught us that no problem is insurmountable—it just hasn’t been solved yet.

Staying committed to the project during difficult times was another challenge. We faced opposition, resource constraints, and months of stalled progress. But our desire to make a real impact kept us going. We set realistic yet ambitious goals, sought advice and alternate perspectives, and maintained our focus. Most importantly, despite the ease of abandoning the project, we never gave up. We might set SENSE aside for a while, but we remained obsessed with seeing it through.

Our dedication was a double-edged sword that propelled us up to this point but also made us fall in love with our idea. The more we invested in the project, the harder it became to assess it objectively. Initially, we defended our idea from criticism, but over time, we learned that scrutinizing our work is essential for improvement. We now strive to evaluate SENSE from different perspectives, staying open to change and focusing on what’s needed rather than what we want.

Balancing our project with other commitments was also a challenge. Yeonjae is doing internship, and I’m completing my final year project. We to make the most out of our limited time, we used tools like meeting notes, calendar events, and to-do lists, prioritizing tasks and setting realistic deadlines.

We expect more challenges ahead, but we welcome them, knowing they will make us stronger.

Reflection

Through this journey, I’ve gained numerous skills and insights:

• Circuit Design
• Data acquisition
• Signal processing and transformation
• Machine learning (CNN)
• Medical device requirements and considerations
• There are only two type of problems: problems that have been solve, or problems that are yet to be solved
• Nothing is given before you try, so do not give up without even trying
• Subtle balance between meticulous preparation and audacious execution is extremely important in any project
• Never make compromise with yourselves
• Good work can come from individuals, but great work can only come from teams
• Everything is hard, scary, and seemingly impossible at first, but nothing stays as your first
• Not falling in love with your own idea
• Habit of always questioning myself to steer in the right direction
• Thinking in different perspectives
• How to plan, propose, and initiate research

Initiating this project has been a transformative experience. What started as a dream has grown into a reality, teaching me that anyone with dedication and passion can achieve something significant. With the wealth of information available online, anyone can learn and create.

Starting something new has always been daunting, but this project has shown me that while the fear is ephemeral, while the skills and results endure. Working with Yeonjae has helped me balance meticulous planning with timely execution. His approach of starting first and adjusting as we go complements my tendency to perfect the plan before beginning. This balance, along with the power of teamwork, has been crucial to our progress. I also learned that having a team to share thoughts, motivate each other, and rely on is key to any success.

SENSE has been demanding, often leaving us exhausted, but the lessons we’ve learned and the progress we’ve made keep us going. Amidst all the uncertainty, one thing is clear: we will never give up.

Files

You may find all our code in github.