Moving forward with movement sensors for ataxia

How do you know a new treatment is working? This is the key question that clinicians have to answer when developing and testing new therapies. One on one, a patient and doctor can discuss symptoms, track improvements in movement and coordination, and decide on future directions together. But how do you know if a new treatment is working across many different doctors and patients? Clinical trials constantly seek out ways to understand how treatments affect an entire patient population. Confidence in results requires consistent ways to measure and assess outcomes across a range of clinics and patients. To account for this variability, clinicians use tests like the International Cooperative Ataxia Rating Scale (ICARS), Scale for the Assessment and Rating of Ataxia (SARA), and Brief Ataxia Rating Scale (BARS), which describe specific movements for doctors to observe and score. These clinical tests are designed to set a standard of assessment for clinicians and consistently represent the progressive worsening of symptoms with heritable ataxias.  

Clinical scoring is effective, but there are ways to innovate this system. Some aspects of note include: 

1. Clinical visits required. Because clinical scoring requires clinician insight, regular monitoring as part of a clinical trial requires regular visits with a doctor or research staff. This can put additional burden on patients with ataxia and their families. Developing strategies for reliable home testing reduces this travel burden and improves clinician’s ability to track patients over time. 
2. Variability between clinics and clinicians. Because scores are ultimately dependent on the observer doing the scoring, this introduces potential variability between clinics and clinicians. This could depend on the observer’s familiarity with ataxia or their recent experiences and observations. Developing a way to measure changes in movement and coordination that doesn’t rely on observer insight may help to reduce variability in collected data. This could help clinicians tell if a potential treatment is working. 
3. Scored tests create artificial floors and ceilings. Have you ever been asked to rate something from 1 to 10 and been tempted to say 11? Or 0? If so, you can see that a key drawback in numerical ranking is the assignment of artificial caps. These are called artificial “floors” for the low end of the range and “ceilings” for the high end of the range. Clinical scoring systems make their systems broadly applicable by having clearly written definitions for each score on the scale, but by doing this they collapse the full extent of movement and coordination changes into a limited range that can be maxed out. Developing a tool to measure movement and coordination changes in a continuum, i.e. with no set score ranks and no caps, allows for more nuanced tracking of a patient’s changing symptoms over time and also removes any arbitrary maximums from scoring.  

In 2021, Dr. Gupta’s lab developed and tested a wearable wrist device that tracked movement during a finger-to-nose test. The device works like an accelerometer, measuring changes in direction and speed in three dimensions. They found qualities of the sensor’s movement during the test (referred to as “submovements” in this and other papers from this lab, see the glossary for a more detailed description) correlated well with a patient’s BARS score. This suggests that this tool could be useful as a supplement for current scoring practices or for symptom monitoring outside of the clinic (Oubre et al 2021). While this was promising, that study only assessed the wrist sensor with one movement test (finger-to-nose). This left several questions that the researchers wanted to address. First, are some movement tests more sensitive than others? For example, how different is movement data from a finger-to-nose test when compared to a different arm movement or even a leg movement? Then, with future clinical trials in mind, how should movement data from multiple unique movement tests be combined to best represent the severity of a patient’s ataxia? 

To answer these questions, the researchers asked study participants to wear movement sensors to track arm and leg movement during five established movement tests. This included the finger-to-nose, finger-chase, heel-stomp, heel-shin, and fast alternating hand movements (AHM) tasks. Participants also completed the BARS test to get a clinical score of ataxia severity. 27 participants without ataxia and 70 participants with ataxia were included in the study, representing a range of ages with an average age of 41 years old in the former group and 57 years old in the latter. Participants with Ataxia represented over 30 different diagnoses, including spinocerebellar ataxia (SCA) types 1, 2, 3, 6, 7, 8, and 14, Ataxia Telangiectasia (A-T), Autosomal recessive ataxia (ARCA) 1 and 3, Multisystem Atrophy – Cerebellar type (MSA–C), Hereditary Spastic Paraplegia (HSP), and others. This inclusion is useful as it helps demonstrate how representative these wrist sensor results might be for different underlying causes of ataxia.  

They compiled movement data from all 5 tests and observed how the main components of movement for each task compared to one another (For example, comparing the finger movement from nose to target versus from target to nose in finger-to-nose task or up versus down movement of the ankle in the heel stomp task). The researchers found that different limbs (wrist versus ankle) and different movement tasks on the same limb (finger-to-nose versus finger chase) result in movements that are mathematically distinct. Interestingly, the type of movement was a better distinguishing feature than ataxia severity. This means that when we’re comparing the submovement profiles from each of the 5 tested movements for participants with and without ataxia, the difference in wrist movement between the finger-to-nose and finger chase tests is more apparent than the effect of ataxia. This suggests that ataxia affects each movement test in a slightly different way. As such, a broad overview of the data might not be able to clearly distinguish small changes in ataxia progression. The researchers then dug into the characteristics of the submovements for each task. They found that there were specific characteristics for each task that appeared to be more affected in participants with ataxia. This demonstrates that the movements of different limbs change in unique ways when a participant has ataxia.  

The impact of ataxia on different submovement characteristics seems to be specific to the limb and test type. Because of this, combining movement data from across different movement tasks will require a more nuanced approach than just averaging or adding values to make a single metric. To accomplish this, the researchers developed a mathematical model that was able to consider both the type of task and the specific submovement within that task to predict the severity of a participant’s ataxia. They tested the reliability of this model against clinical BARS scores and found that it correlated well. This suggests that the movement sensor data could be used to measure ataxia severity. 

In summary, the newest paper from the Gupta Lab (Patel et al 2025) builds on previous work studying the usefulness of wearable movement sensors in assessing ataxia severity. They had study participants perform 5 movement tests commonly used in the clinic that require coordinated arm or leg use. They found that ataxia caused unique changes in movement for each test. This suggests that ataxia differently affects different limbs or movements within the same limb.  

This isn’t an entirely surprising finding. Though the whole body experiences reduced coordination in ataxia, the gestures performed during the tested movements use different muscle groups which require unique brain inputs to coordinate them. To best represent ataxia severity using the sensor data, the researchers developed a mathematical model that takes movement data from the sensor and combines it with information about the type of movement tested. The predictions from this model correlate well with clinician BARS scores. This suggests that this model can be used with the sensors for future research tracking ataxia severity either as a part of a clinical trial. It could even be a way for clinicians to track ataxia progression through at-home monitoring. Dr. Gupta and his team are continuing to research how wearable movement sensors could be used for ataxia symptom monitoring.