Leveraging Mobile Technology and Machine Learning for Enhanced Visual Function Assessment in Inherited Retinal Diseases

PhD study funded by Graduate Teaching Assistantship (GTA) studentships in Computer Science

Highlights

1. Implementing an innovative, smartphone-based AI eye tracking system to facilitate early detection and continuous monitoring of inherited retinal diseases (IRDs).
2. Design a mobile application that facilitates self-monitoring and management of hereditary retinal dystrophies (IRD) by patients in a home setting, using advanced smartphone technology to enhance inclusivity and awareness of the condition, thereby aiming to decrease the frequency of clinical consultations within the NHS and associated healthcare expenditures.
3. This project brings together expertise across domains MDE, ML and clinical ophthalmology and leverages access to a tertiary clinical centre for IRD. This is reflected by the strong publication track record in IRDs and the recent 4* REF impact case study “Improving Diagnosis and Patient Outcomes in Developmental Retinal Diseases (DRDs)”. The outcomes of this project will contribute to the impact case study “AI and its applications” at CMS.

Project

Navigating the world around us requires more than just clear vision; it demands the ability to make rapid and precise eye movements in response to our ever-changing environment. This seamless interaction with our surroundings is made possible by a well-functioning visual system that keeps our gaze steady and fixed on what we choose to see. However, for individuals living with Inherited Retinal Diseases (IRDs), this fundamental aspect of vision is compromised. IRDs are a group of genetic disorders that affect the retina’s ability to detect light and colour, leading to progressive vision loss. Patients with these conditions often experience difficulties in tracking moving objects, stabilizing their gaze, and adapting to different lighting conditions, which can make everyday tasks challenging and diminish their quality of life.

The impact of IRDs extends beyond the physical limitations of vision loss. The constant struggle to see clearly can lead to social isolation, difficulty in education and employment, and a significant emotional toll. Traditional assessments of retinal health and eye movement are typically conducted in specialized settings by experienced clinicians, which can be costly, time-consuming, and inaccessible for many. Moreover, these evaluations may not capture the full extent of the patient’s daily challenges or the progression of their condition over time. There is a pressing need for innovative solutions that can provide accurate, accessible, and continuous monitoring of eye 

health and function for those affected by IRDs. By leveraging the capabilities of smartphones, we have an opportunity to bridge this gap, offering patients a tool to better understand and manage their condition, ultimately improving their independence and well-being. Additionally, this project aims to reduce the costs associated with clinical visits by providing a reliable eye-tracking tool that can be used at home.

This project aims to improve IRD care by investigating mobile health (mHealth) technology that employs eye-tracking technology and machine learning to offer a nuanced, personalized assessment of visual function in IRD patients without the need of frequent clinical consultations. This initiative not only seeks to support individuals with IRDs but also lays the foundation for a platform that could be expanded with further funding to develop tools that assist a broader range of patients, enhancing support and care for those with visual impairments.

Research objectives:

1. To develop an effective data collection framework that utilizes smartphone sensors (front-facing camera, ambient light sensor, and proximity sensor) for gathering high-quality data on gaze direction, eye movement, and pupillary responses under various lighting conditions.
2. To conduct a comparative analysis of common (hereditary) retinal dystrophies and standard vision tests (e.g., Fixation Stability Test, Smooth Pursuit Test, Pupillary Light Reflex) to identify specific eye movement patterns and anomalies associated with these conditions.
3. To design, train, and evaluate a hybrid machine learning model that effectively combines Convolutional Neural Networks (CNN) for analysing eye images and Recurrent Neural Networks (RNN) for interpreting the temporal sequence of eye movements, aiming for high accuracy in eye gaze and movement detection.
4. To integrate the eye-tracking algorithm into a user-friendly mobile app, building a proof-of-concept of the technology.

Methodology:

• Data collection framework: Implement software algorithms to process input from the smartphone’s front-facing camera, ambient light sensor, and proximity sensor. This involves developing calibration techniques to personalize the eye-tracking system for individual users, accounting for variations in IRD, eye anatomy and visual acuity.
• Comparative Analysis: Review existing literature and collaborate with ophthalmologists to understand the progression of retinal dystrophies and their impact on eye movements, and the information that the tests can bring to the patients. This will guide the selection of relevant vision tests and the identification of key data points for machine learning model training.
• Machine Learning Model Development: Utilize a dataset of eye movement recordings to train a hybrid CNN-RNN model. The CNN component will analyse static images for features like gaze direction and pupil size, while the RNN component will process sequences of images to capture dynamic eye movements.
• Evaluation: Conduct a pilot study with participants diagnosed with hereditary retinal dystrophies. The study will assess the app’s accuracy in tracking eye movements, its effectiveness in monitoring disease progression, and its usability in everyday scenarios. Feedback from participants will inform iterative improvements to the app. The project’s success will be evaluated through several metrics: the accuracy of the eye-tracking system compared to traditional methods, the machine learning model’s performance in identifying eye movement anomalies, and user satisfaction with the eye tracking technology on the mobile app. A combination of quantitative data analysis and qualitative feedback will help understand the app’s effectiveness on monitoring (hereditary) retinal dystrophies.

Enquiries to Dr Artur Boronat [email protected] or [email protected]

Further details and application advice at https://le.ac.uk/study/research-degrees/funded-opportunities/cms-gta