Artificial Intelligence enables researchers to accurately detect signs of Parkinson’s disease by analyzing facial muscles. At the University of Rochester, Ehsan Hoque and his team have developed computer vision software to predict an individual’s likelihood of developing Parkinson’s disease by analyzing their selfie photos.
Daily, countless individuals capture selfies using their smartphones or webcams, eagerly sharing them online. It’s a universal trend, and it’s almost impossible not to notice the smiles on their faces.
But to Ehsan Hoque and his team at the University of Rochester, these photos hold immense value beyond mere visual content. Their developed computer vision software, equipped with advanced algorithms, can analyse these brief videos, including the snippets recorded during selfie moments. It possesses the remarkable ability to detect subtle facial muscle movements that remain imperceptible to the naked eye.
With extraordinary accuracy, this software can predict whether individuals frequently take selfies are prone to developing Parkinson’s disease. According to Hoque, an associate professor of computer science at the University of Rochester, “Parkinson’s is the fastest-growing neurological disorder.
“WHAT IF, WITH PEOPLE’S PERMISSION, WE COULD ANALYZE THOSE SELFIES AND GIVE THEM A REFERRAL IN CASE THEY ARE SHOWING EARLY SIGNS?”
The researchers’ technology is described and published in a study titled “Facial expressions can detect Parkinson’s disease: preliminary evidence from videos collected online” recently in the journal Nature Digital Medicine by a team from the University of Rochester.
The research team conducted a study where they analysed a total of 1,812 videos. These videos featured 604 individuals, with 543 not having Parkinson’s disease and 61 individuals diagnosed with Parkinson’s disease. The participants had an average age of 63.9 years and were asked to perform various tasks online while being recorded by a webcam.
During the test, patients are engaged in a series of actions that delve deep into their facial expressions. These actions demand their utmost involvement, including:
- Articulate a complex written sentence out loud.
- Swiftly tap their index finger to their thumb, aiming for maximum speed for ten repetitions.
- Evoke the most robust expression of disgust, alternating it with a neutral expression and repeating this cycle thrice.
- Ascend their eyebrows to the highest point they can reach, then descend them to their lowest position, gradually repeating this movement three times.
Using advanced machine learning algorithms, the computer program swiftly provides a percentage likelihood, based on each test, indicating whether the patient exhibits symptoms of Parkinson’s or related disorders.
But what does the program specifically analyse? When patients smile, the software examines their control over facial muscles, mainly focusing on a symptom known as “modularity” in the medical field, which refers to decreased muscle coordination associated with Parkinson’s.
“One thing about Parkinson’s is that you don’t show all the symptoms all the time, and not every symptom is shown in every part of your body,” explains Rafayet Ali, a former postdoctoral associate in Hoque’s lab who now is an associate data scientist at Sysco, “For example, you may not have hand tremors, but you may show a significant level of deviation in your smile.”
Transitioning from conventional pen-and-paper assessments to modern “objective, digital evaluations.”
Hoque and Ali have a personal connection to assisting individuals with Parkinson’s disease. Both of their mothers have been affected by the disorder. Hoque’s late mother in Bangladesh, for instance, underwent treatment with levodopa, a leading medication for Parkinson’s, after finally connecting with one of the country’s limited number of neurologists. This treatment successfully alleviated her tremors, bringing immense happiness to their family. However, due to challenges in scheduling follow-up appointments, the earthquakes eventually resurfaced.
Motivated by these experiences, Hoque reached out to Ray Dorsey, a professor of neurology, in an informal email exchange. When they finally met in 2016, Dorsey presented Hoque with a comprehensive document containing the forms required for the Movement Disorders Society–Unified Parkinson’s Disease Rating Scale (MDS-UPDRS). This rating scale is the gold standard for assessing Parkinson’s disease and is traditionally conducted using pen and paper. Dorsey expressed his enthusiasm for incorporating automation and data analytics into the process, recognising the potential contributions they could make.
According to Dorsey, “Objective, digital assessments of Parkinson’s disease can help us diagnose people with the condition and evaluate new therapies for the condition faster” However before Hoque and his team can seek permission to analyse people’s selfies or implement the five-pronged test they have developed, extensive precautions must be taken. Hoque emphasises the need for caution and adherence to FDA guidelines to ensure the accuracy and safety of the algorithm when used by individuals worldwide.
The Path to FDA Approval
While significant progress has been made in automatically analysing facial expressions, voice, and motor movements to detect Parkinson’s disease, further research is necessary to develop algorithms to distinguish these involuntary tremors from those associated with other movement disorders, such as ataxia and Huntington’s disease. Hoque acknowledges that this differentiation has yet to be achieved, but the pursuit of utilising artificial intelligence to prevent misdiagnosis and maximise benefits remains ongoing.
The research conducted by the University of Rochester team is progressing with the generous support of a $500,000 grant from the Gordon and Betty Moore Foundation.
Is Speech the Key?
People increasingly utilize voice-activated intelligent devices like Apple Watch and Google Voice Assistant to complete daily tasks. Can these devices analyze our speech and voices and alert us to early warning signs of Parkinson’s disease?
Recent research conducted by Rochester scientists suggests that it is possible in a study published in the Journal of Medical Internet Research, researchers such as Wasifur Rahman, Sangwu Lee, Md. Saiful Islam and others in Hoque’s laboratory demonstrate how an online tool can remotely screen individuals for Parkinson’s disease using speech tasks enabled by video or audio.
Collectively, these research efforts contribute to a future where access to neurological care is as widespread as owning a smartphone or any other internet-enabled device.
It will take giant technological strides, but wait, there is an option…
Visual and audio biomarkers hold immense potential for detecting and monitoring Parkinson’s disease. However, it is essential to acknowledge that clinical validation and FDA clearance are crucial steps for these technologies to be widely implemented in an affordable and user-friendly manner within the Parkinson’s community. While these processes are still underway, existing options harness the power of wearable technologies and artificial intelligence for passive monitoring of Parkinson’s.
Connecting Technological Advancements to Care: The Role of Parky
While these technological advancements hold immense potential to improve care in Parkinson’s, further research is needed. However, there are validated and FDA-cleared ways to enhance healthcare and self-care in Parkinson’s through self-tracking of symptoms via Apple Watch. Parky, a groundbreaking doctor-prescribed digital health tool that empowers individuals to monitor their motor symptoms, enhance medication adherence, and facilitate seamless communication with healthcare professionals. With Parky’s objective data source, patients and healthcare providers get support for optimal treatment plans.
Parky’s Capabilities:
- Motor Symptom Tracking: Using Apple Watch, Parky enables individuals to monitor motor symptoms accurately and efficiently.
- Medication Adherence: The digital health tool assists individuals in sticking to their prescribed medication regimens, promoting better management of Parkinson’s symptoms.
- Data Exchange: Parky fosters seamless communication between patients and medical professionals, ensuring that crucial data informs informed decision-making.
How to use Parky for Parkinson’s app?
For individuals seeking to incorporate Parky into their Parkinson’s care routine, the process is seamless. Parky is a prescription-only app, meaning that healthcare providers can log in to the Parky HCP Portal and prescribe Parky online within seconds. Once prescribed, users can download the app from the App Store and start using it. The only requirement is to wear their Apple Watch, and the rest is passive and automatic. For detailed information on how to get started, visit the Parky for Users.
Embracing Technological Strides for Parkinson’s Care
In the pursuit of transforming Parkinson’s care, the combination of AI-driven facial analysis, speech screening, and innovative tools like Parky holds immense promise. The ongoing collaboration between researchers, the cautious navigation of FDA approvals, and the integration of technological innovations exemplify a collective commitment to revolutionizing the landscape of Parkinson’s disease detection and management.
