Title: Impact of Video Quality on Video-based Digital Assessment of Parkinson’s Disease

Authors: Atefeh Irani, Kye Won Park, Maryam S. Mirian, Mahdi Shakouri, Michael Grundy, Juana Ayala, Reshad Hosseini, Hadi Moradi, Martin J. McKeown

Summary

Remote video analysis is increasingly used to measure Parkinson’s disease motor symptoms at home. These recordings allow clinicians and researchers to track symptoms such as slowed or reduced movement more frequently, and in some cases from a patient’s own home. This approach, often called using “digital biomarkers,” has the potential to make monitoring more convenient and more reflective of everyday life. However, home recordings are not always captured under ideal conditions, and differences in lighting, camera setup, or internet limitations may affect video quality.

In this study, our researchers examined how video quality influences the accuracy of automated Parkinson’s assessments. We analyzed finger-tapping videos collected both in clinic settings and in real-world home environments, and systematically tested how factors such as video resolution, frame rate, lighting, and camera positioning affected the system’s ability to track hand movements and estimate clinical scores.

We found that two factors were especially important for reliable results: recording speed and clear visibility of the hand. Videos recorded at fewer than 24 frames per second significantly reduced measurement accuracy, and recordings where the hand appeared too small or too large within the frame also increased errors. Encouragingly, extremely high video resolution was not necessary, even lower-resolution videos produced reliable results when recording speed and positioning were appropriate.

These findings help establish simple, practical guidelines for patients recording videos at home, supporting accurate and clinically meaningful remote assessments. By identifying the key requirements for reliable recordings, this work helps advance safe and scalable video-based monitoring of Parkinson’s disease, bringing symptom tracking closer to patients’ everyday environments while maintaining clinical quality.

Title: Facial expression analysis to uncover the relationship between sialorrhea and hypomimia in Parkinson’s disease

Authors: Eline Serbée, Kye Won Park, Atefeh Irani, Maryam S. Mirian, Juana Ayala Castaneda, Michael Grundy, Martin J. McKeown

Link: https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2025.1661043/full

Summary:

Drooling (medically called sialorrhea) and reduced facial expression (hypomimia) are two features commonly seen in Parkinson’s disease (PD). While both can affect quality of life, they are often under-recognized in clinic. This study looked at how these two symptoms relate to each other using both traditional clinical scales and automated video-based facial analysis.

Researchers recorded short videos of facial expressions from 52 people with PD and used artificial intelligence to measure specific features of facial movement around the mouth and eyes. They also used a validated questionnaire to assess how much drooling individuals experienced. Clinical rating scales showed a modest link between reduced facial expression and more severe drooling. When facial video features were analyzed with computer algorithms, the relationship was stronger with certain patterns in facial movement correlating moderately with drooling severity.

These findings suggest that automated facial analysis can help quantify the relationship between facial expressivity and drooling in PD. Because facial videos are easy to record, this approach could improve detection of non-motor symptoms that are otherwise hidden or under-reported. Ultimately, better recognition of these symptoms may lead to more comprehensive symptom monitoring and targeted care strategies for individuals living with Parkinson’s disease.

Title: HiLWS: A Human-in-the-Loop Weak Supervision Framework for Curating Clinical and Home Video Data for Neurological Assessment

Authors: Atefeh Irani, Maryam S. Mirian, Alex Lassooij, Reshad Hosseini, Hadi Moradi, Martin J. McKeown

Summary:

Video recordings are increasingly being used to help monitor movement symptoms in people living with Parkinson’s disease, allowing clinicians to assess changes more frequently and sometimes from patients’ homes. However, videos recorded outside the clinic often vary in quality due to differences in lighting, camera setup, or how tasks are performed, which can reduce the accuracy of automated assessments. In this study, researchers developed a new framework called HiLWS (Human-in-the-Loop Weak Supervision) that combines artificial intelligence with expert clinical input to improve how medical videos are prepared, labeled, and analyzed for remote neurological assessment.

Using more than 2,000 videos collected in both clinical and home environments, the team showed that real-world recording conditions significantly influence measurement accuracy. They identified practical guidelines for reliable recordings, including maintaining adequate recording speed and ensuring the hand is clearly visible in the frame, while demonstrating that very high video resolution is not required. The framework also improved consistency in symptom scoring by combining input from multiple clinical experts with machine learning models and directing expert review toward uncertain cases.

These findings help make remote Parkinson’s assessments more reliable and scalable, supporting more frequent monitoring outside the clinic while maintaining clinical quality. More broadly, the approach provides a foundation for building trustworthy AI tools for remote healthcare, helping expand access to neurological care and enabling continuous monitoring for people living with chronic conditions.

Title: Low-Intensity Transcranial Ultrasound Effects on the Ventral Intermediate Nucleus and Zona Incerta for Parkinson’s tremor

Authors: Maggie Q. Vuong, Varsha Sreenivasan, Soojin Lee, Ben Quinn, Martin S. Keung, Michael Grundy, Juana Ayala Castañeda, Hongchae Baek, Martin J. McKeown

Link: https://www.sciencedirect.com/science/article/pii/S1935861X26000021

Summary:

Tremor is a common and challenging symptom of Parkinson’s disease (PD) that does not always respond well to medication. While high-intensity focused ultrasound is an FDA-approved method used to reduce tremor by targeting a small brain region called the ventral intermediate nucleus (VIM), Parkinson’s tremor is complex because it can occur both at rest and while holding a posture. Another nearby brain region, the zona incerta (ZI), is also linked to tremor and connected to broader movement networks. In this study, we used low-intensity focused ultrasound (LIFUS), a non-invasive and temporary form of brain stimulation, to better understand how modulating brain activity in these two regions affects tremor in people living with PD.

Nineteen participants with PD took part in the study. Each person first underwent MRI scanning to precisely guide ultrasound targeting. On the same day, participants received two separate LIFUS sessions: one targeting the VIM and one targeting the ZI. Before and after each session, participants completed tasks designed to bring out tremor while researchers measured movement using small motion sensors placed on the arm. Functional MRI scans were also collected to examine how brain activity changed following each type of stimulation.

The results showed differences between the two brain targets. Stimulation of the VIM reduced tremor mainly when participants were holding a posture, which is consistent with previous research. In contrast, stimulation of the ZI mildly improved both postural tremor and tremor at rest, suggesting it may influence a wider network involved in movement control. Brain imaging supported this finding, showing broader changes in brain activity after ZI stimulation compared with VIM stimulation. Together, these findings suggest that while the VIM remains an effective target for certain types of tremor, the ZI may represent a promising new target for treating Parkinson’s tremor more broadly in the future.

Title: Misfolded protein deposits in Parkinson's disease and Parkinson's disease-related cognitive impairment, a [11C]PBB3 study

Authors: Michele Matarazzo, Alexandra Pérez-Soriano, Nasim Vafai, Elham Shahinfard, Kevin Ju-Chieh Cheng, Jessamyn McKenzie, Nicole Neilson, Qing Miao, Paul Schaffer, Hitoshi Shinotoh, Jeffrey H Kordower, Vesna Sossi, A Jon Stoessl

Link: https://www.nature.com/articles/s41531-024-00708-z

Key Findings
Misfolded Protein Aggregation in PD
One of the study’s significant findings is the higher [11C]PBB3 binding in an area of the brain called the posterior putamen of PD subjects compared to healthy controls. This indicates an accumulation of misfolded proteins in this region, highlighting it as a critical site of protein aggregation in Parkinson’s disease. However, there was no significant difference found in [11C]PBB3 binding between PD subjects and healthy controls in another region, the substantia nigra, which is traditionally considered a primary site of dopaminergic neuron loss in PD.    
Cognitive Impairment and Anterior Cingulate
The study revealed that CI-PD subjects show increased [11C]PBB3 binding compared to CN-PD and HC in another area of the brain, the anterior cingulate cortex. This higher binding indicates that greater misfolded protein deposition in this region is associated with more severe cognitive impairment. The results suggest that pathology in the anterior cingulate could serve as a significant marker for cognitive decline in PD, offering potential diagnostic and therapeutic targets for managing cognitive impairments in PD. 
Neuroinflammation
The study also examined the relationship between misfolded protein deposits and neuroinflammation by comparing [11C]PBB3 and [11C]PBR28 binding in nigrostriatal regions. The results show no significant correlation between these two measures, suggesting that the neuroinflammation detected by [11C]PBR28 does not directly relate to the accumulation of misfolded proteins measured by [11C]PBB3. This finding highlights the complexity of PD pathology, where misfolded protein accumulation and neuroinflammation may contribute to the disease through distinct and perhaps independent mechanisms. 

Broader Implications
The findings of this study have several important implications for understanding and treating Parkinson’s disease and its cognitive impairments:

Role of Striatal Axonal Terminals
The preferential involvement of striatal axonal terminals, particularly in the posterior putamen, in misfolded protein aggregation emphasizes the need to consider these regions in PD research and treatment strategies. This challenges the traditional focus solely on the substantia nigra and underscores the complexity of PD pathology. 
Cognitive Decline and Anterior Cingulate Pathology
Identifying the anterior cingulate cortex as a critical area for cognitive impairment in PD provides a potential biomarker for diagnosing cognitive decline. This could lead to earlier and more accurate identification of patients at risk for cognitive impairment, allowing for earlier interventions and better management of cognitive symptoms. 
Therapeutic Targets:
The study's insights into the specific regions involved in misfolded protein aggregation open new avenues for therapeutic development. Targeting the posterior putamen and anterior cingulate cortex with treatments aimed at reducing misfolded protein deposits could help slow disease progression and mitigate cognitive decline. 
Diagnostic Imaging
The use of [11C]PBB3 PET imaging as a tool to visualize and quantify misfolded protein deposits represents a significant advancement in PD research. This imaging technique could be used to monitor disease progression, evaluate the efficacy of new treatments, and enhance our understanding of the underlying mechanisms of PD and its cognitive impairments. 

Conclusion
The study adds to our understanding of how protein clumps affect PD and its cognitive effects. By identifying the posterior putamen as a major site of clumped proteins and showing the link between the anterior cingulate cortex and cognitive decline, the study highlights how different brain regions are involved in PD. It also shows that inflammation and protein clumps might not always go hand in hand, revealing the complexity of PD. PET imaging with tracers like [11C]PBB3 are invaluable for studying these issues, opening the door to better diagnostics and treatments in the future. 

Title: Individualised, brain-informed motor vigour assessment for targeted non-invasive brain stimulation in Parkinson’s disease

Authors: Varsha Sreenivasan, Martin S. Keung, Soojin Lee, Hanaa Diab, Wyatt Vercher, Martin J. McKeown

Summary

A common characteristic of slowness of movement (bradykinesia) in Parkinson’s disease (PD) is reduced motor vigour — the intensity, speed, and motivational drive that propel our movements. Current treatments, including deep brain stimulation and medications such as levodopa, primarily aim to improve bradykinesia. However, these treatments are usually evaluated using simple behavioural measures like reaction time or movement speed. These single measures vary markedly across individuals, and using one uniform measure across individuals is not optimal. Motor vigour is associated with the subjective motivational drive to move and is therefore an individual-specific attribute. In this work, we introduce an individualised, brain-based way of measuring motor vigour and test whether it can perform better than simple behavioural measures in detecting the effects of brain-modulating treatments in Parkinson’s disease.

We studied 18 people with Parkinson’s disease who performed a rapid, forceful hand-squeezing task. During the task, participants received electrical vestibular stimulation. Nine types of stimulation were tested, including a sham (placebo-like) condition. At the same time, brain activity was recorded using functional MRI. Participants performed this task once each before and after taking levodopa medication. We measured six aspects of behaviour, including reaction time, how long the squeeze lasted, and how quickly it was performed. We then analysed patterns of brain activity to identify three brain networks and examined how their responses changed over time. By combining brain activity and behavioural data for each individual using advanced statistical modelling, we created a personalised measure called individualised, brain-informed motor vigour (IndMV).

We show that traditional behavioural measures did not show a clear benefit of electrical stimulation compared to sham. Moreover, these measures could not distinguish between medication states very well (60–82%). On the other hand, the personalised IndMV measure outperformed the traditional behavioural measures by capturing significantly greater changes caused by stimulation and differentiating between medication conditions with 93% accuracy. When stimulation was optimised for each individual, motor vigour improved by more than 30%. IndMV is a new, personalised way of measuring motor vigour that goes beyond standard behavioural tests. By integrating brain and behavioural data, it more effectively captures the effects of brain modulation and offers a promising tool for better identifying and tailoring treatment effects in Parkinson’s disease.

Title: Rapid Profiling of EEG Responses to Non-Invasive Brain Stimulation in Parkinson’s Disease: A Biomarker-Driven Screening Framework

Authors: Sepideh Hajipour Sardouie, Mahdi Babaei, Mahsa Naseri, Servin Mehrtash, Mohammad Hossein Faramarzi, Zahra Kavian, Martin Keung, Varsha Sreenivisan, Hanaa Diab, Maryam S. Mirian and Martin J. McKeown

Link: https://www.mdpi.com/2227-9059/14/2/352

Summary:

Parkinson’s disease affects not only movement but also the brain’s natural rhythms. These rhythms can be measured using electroencephalography (EEG), a non‑invasive method that records electrical activity from the scalp. In recent years, researchers have become increasingly interested in whether gentle forms of brain stimulation can shift these rhythms towards a healthier pattern. However, finding the right stimulation settings for each person remains a major challenge.

In this study, we developed a fast and systematic way to test many different types of Galvanic Vestibular Stimulation (GVS) signals, over 300 in total, while recording EEG in people with Parkinson’s disease and in healthy volunteers. GVS uses very small electrical currents applied behind the ears to subtly activate the balance system, which in turn influences brain activity.

To understand how each stimulation pattern affected the brain, we created a combined “biomarker score” that captured several important features of EEG activity, including brainwave power, interactions between different brainwaves, and signal complexity. Instead of relying on movement or behavioral changes, this score allowed us to directly measure how close the brain’s response moved toward a healthier profile.

We found that several types of stimulation, particularly specific rhythmic patterns, consistently shifted EEG activity in a positive direction. Importantly, these effects occurred only during stimulation, with no lingering changes afterward. Although this was a pilot, exploratory study, the approach shows promise as a rapid screening tool for identifying personalized stimulation strategies. By focusing on brain biomarkers rather than symptoms alone, this framework may help guide the development of more targeted, individualized brain-modulating therapies for Parkinson’s disease.

Title: Comprehensive Benchmarking of EEG Biomarkers for Parkinson’s Disease: Robustness, Reliability, and Clinical Discriminability Across Multiple Datasets

Authors: Sepideh Hajipour Sardouie, Mahsa Naseri, Mahdi Babaei, Servin Mehrtash, Mohammad Hossein Faramarzi, Zahra Kavian, Maryam S. Mirian and Martin J. McKeown

Summary:

Parkinson’s disease affects millions of people worldwide, yet diagnosing and tracking its progression remains challenging. One promising tool is electroencephalography (EEG), a simple, non‑invasive method that measures the brain’s electrical activity. Many research groups have proposed EEG‑based “biomarkers” for Parkinson’s, but their findings often disagree. This inconsistency makes it difficult to know which brain signals truly reflect the disease and which are too sensitive to differences in data processing or analysis.

Our study takes an important step toward solving this problem. We performed the first large, systematic comparison of 16 commonly used EEG biomarkers across two independent datasets of people with Parkinson’s and healthy adults. We tested how stable each biomarker is when researchers change key analysis choices, such as how much denoising is applied, how long the data segments are, or how signals from multiple electrodes are combined. We also examined how reliably these biomarkers can distinguish Parkinson’s patients from healthy individuals, and how well they capture changes when patients are on or off their medication.

We found that some biomarkers, especially those describing the brain’s background activity (aperiodic features), as well as measures of signal complexity and variability, were consistently robust and clinically informative. In contrast, biomarkers based on rhythmic brain activity were much more sensitive to methodological choices and less reliable in recordings of the brain at rest. By identifying which EEG features are truly dependable, our work provides practical guidance for researchers and clinicians. This benchmarking framework can help standardize EEG analysis and accelerate the development of reliable brain‑based markers for diagnosing Parkinson’s, monitoring symptoms, and evaluating new treatments.