Page 175 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

2 Use Case Description

2�1 Description

Facial paralysis is a condition in which the patient is unable to control the muscles of their 4.1-Healthcare
face. While some causes, such as Bell’s palsy, may resolve spontaneously, others, particularly
neurological conditions, require early detection to prevent the progression of symptoms and
enable timely medical intervention. Identifying the early signs of facial paralysis can serve
as an indicator of neurological disorders such as strokes and multiple sclerosis; infectious
conditions such as Bell’s palsy, Lyme disease, and Ramsay Hunt syndrome; physical trauma,
including injuries from surgery or fractures; the presence of tumours; and other specific
disorders such as Guillain-Barré syndrome. [1] We propose the use of a wearable, video-based
monitoring system inspired by the design of a smartwatch, which takes advantage of the fact
that such a device is naturally directed towards the user’s face multiple times throughout the
day. Recognising that many patients may not regularly interact with dedicated software, the
system is designed to continuously and passively monitor the user’s face whenever the watch
is turned towards them. Each time the patient raises the watch to check the time, the system
automatically captures a brief video of the face. Additionally, the patient is expected to spend
approximately one minute per day performing facial mimicry exercises, which allow the AI
system to assess facial muscle functionality. To reduce the need for user interaction, which may
be a deterrent for some patients, the software is designed to automatically detect whether the
patient is engaging in these exercises during the captured footage. This ensures a completely
hands-free experience. The software keeps a log of the mimicry exercises performed and their
respective timestamps, enabling it to prompt the patient to repeat specific gestures if they have
not been performed for some time. The device employs existing mechanisms to detect hand-
raising gestures, commonly referred to as the “flip to wake” or “raise to wake” feature. Captured
videos are transmitted to a nearby fog-layer computational unit, such as a mobile phone,
which subsequently uploads them to cloud servers equipped with specialised deep learning
models for facial paralysis detection. These models, developed within the academic research
community, are sufficient for analysis as the system does not require any bespoke training
per individual patient. However, access to timestamped video history enables the models to
compare current facial data with that of previous instances, potentially improving diagnostic
accuracy by accounting for the patient’s natural facial muscle range and individual baseline.

Use Case Status: Existing solutions [2]:

Partners: Vellore Institute of Technology, Chennai [7]

2�2 Benefits of use case

• Early detection and monitoring of facial paralysis can lead to timely medical interventions,
improving patient outcomes and quality of life.
• Developing new technologies, such as AI-based analysis or remote monitoring
applications, contributes to medical innovation and accessibility.
• Collaboration between researchers, healthcare providers, and technology developers is
essential for implementing effective solutions in facial paralysis diagnosis and treatment.

139