Project Background
Dropped head syndrome (DHS) is defined by the inability of a person to move and raise their head. DHS results from neck muscle weakness which can be caused by neurological impairments like amyotrophic lateral sclerosis (ALS), autoimmune conditions, or exposure to radiation therapies. People with moderate DHS cannot maintain an upright head posture for extended time due to fatigue. In severe cases, the head completely drops, resulting in a "head-on-chest'' posture. DHS of any degree markedly affects all daily functions, and patients tend to withdraw from all activities.
Neck collars are currently used to keep the head upright; however, few patients use them due to their inability to rotate the head and the associated discomforts. Although recent collars added compliance to allow for small voluntary head rotations, the user has to fight against the structural stiffness which is difficult to achieve.
By contrast, wearable robots like neck exoskeletons present an opportunity to enable head movements for these patients. Our lab has contributed significantly to the state-of-the-art of the powered neck exoskeleton technology. As shown in the image, our Utah Powered Neck Exoskeleton allows for sufficient range of motion, intuitive control via a user's gaze, and adaptive fitting for comfort. However, a barrier for clinical translation of this device is the significant cost associated with using a commercial eye-tracker (> $6,000) for gaze sensing.
The objective for this project is to develop a low-cost gaze sensing system. An approach combining mechatronics design and modeling using machine learning will be used. In terms of the outcomes, the accuracy and cost of this system will be compared to a commercial eye tracker.
Student Role
The undergraduate student is expected to work full-time (40 hours weekly) and will be responsible of three main tasks:
(1) Train machine learning models to predict gaze location within the field of view, based on a dataset gathered by using a commercial gaze tracker.
(2) Perform analysis and optimize models for prediction accuracy, as compared to ground truth provided by the commercial tracker algorithm.
(3) Build a mechatronic eye-tracking system that houses low-cost video and infrared cameras and is capable of streaming data in real-time at above 30 Hz.
We have already collected a dataset using a commercial gaze tracker (Neon, Pupil Labs, Berlin, Germany). This commercial tracker has one video camera serving as a proxy of the field of view, as well as two infrared (IR) cameras, each taking images of an eye pupil. In the dataset, we have time-synchronized images from these cameras and the measured gaze location by the black-box gaze tracker algorithm (ground truth).
In the first task, the student will develop machine learning models to predict gaze location in the field of view based on the image inputs from the dataset. The student will need to develop a pipeline to train, test, and validate the machine learning models using the dataset.
In the second task, the student will quantify the accuracy of their models and optimize models to achieve highest prediction accuracy during validation (unseen data).
In the third task, which is independent of the first two tasks, the student will build frames and electronic wirings to package low-cost video and IR cameras. The goal is for the structure to be lightweight and capable of integrating with the existing neck exoskeleton structure.
Student Learning Outcomes and Benefits
The student will have the opportunity to work in a multidisciplinary laboratory. The PI will provide needed research resources and mentorship throughout the program. A graduate student will be assigned to also mentor the student during his/her stay.
The student is expected to learn skills that are relevant to robotics, mechatronics, and design, such as computer programming, 3D CAD and prototyping, and data collection, validation, plotting, and analyses. Additionally, the student will gain opportunity to work with user studies, including movement sensors, data collection and processing, and analysis. This project will help the student learn how to conduct research, how to acquire skills as needed, and how to communicate and collaborate with others in a laboratory setting.
The student will also gain opportunity to present and share his/her results from this novel project through manuscript preparation and conference presentations. The student will be encouraged to complete all programmatic aspects of SPUR, including attending orientation, bi-weekly meetings and weekly URES events, presenting at the Summer Symposium, and publishing an abstract in the University of Utah Undergraduate Research Journal.
The research experience will go beyond traditional classroom model, thus enriching the learning experience of the student. Ultimately, the goal is to prepare and inspire the student through this research experience to pursue a successful STEM career, thereby contributing the nation's STEM workforce.
