Context-aware monitoring of patients with upper-limb neurological disorders in free-living conditions

  • Contact person: Yasaman Izadmehr
  • People involved: Yasaman Izadmehr, Prof. Anderes Perez-Uribe, Prof. Kamiar Aminian
  • Partners: Centre hospitalier universitaire vaudois, Hospital Nestlé du CHUV
  • Funding source:

Loosening picture


The primary step to monitor and guide patients with upper-limb neurological disorders is objectively assessing upper limb movement impairment accurately. Currently, the majority of diagnosis and treatment measures of the patients with upper-limb neuro-motor disorders are based on the subjective assessment performed by the clinical specialist in the hospital based on operator- dependent, ordinal scaling and subject to floor and ceiling effects. Moreover, the medical staff does not have an objective assessment of the patient’s actual performance at home, which is an unconstrained environment. Allowing the continuous monitoring of patients’ activities in free living should provide doctors with unprecedented insights into the patient’s movement impairment.

The objective of this study is to provide a set of technologies, mainly exploiting wearable and embedded sensor devices (e.g., inertial measurement unit (IMU) and an egocentric camera) to monitor and objectively assess the patients’ actual performance in a daily life setting. The objective assessment of the quality of movement will be based on a series of measurements, which will take place both, in the hospital according to a predefined controlled setup, and in uncontrolled conditions.

Constructing a connection between the measurements in the hospital and those in free-living is achievable thanks to the a priori definition of a small set of sequence movements. Those movements are defined by the clinical specialists to assess the patient’s quality of movement. Thus, one of the aims of this research is to come up with objective measures that work with those movements, and moreover to be able to spot those movements (or sub-movements) in free-living. Movement recognition will be achieved using time-series analysis and low-shot Machine learning algorithms processing both, the time-series provided by the wearable IMUs and time-series data captured by a 3D egocentric camera.

In order to allow the interpretation of the objective measures obtained in uncontrolled settings, we will work towards endowing our system with context-aware capabilities by processing the ego- centric videos with state-of-the-art machine learning algorithms (e.g., Recurrent spatial-temporal attention networks, Temporal Convolutional Networks, etc.) for recognizing the activities being performed by the user and providing further contextual information, including where the activities are taking place, the objects the patient is interacting with, etc. Last but not least, we will attempt to enrich the context of the measurements by assessing the current “state” of the patient, characterized for instance, by the readings of diverse wearable physiological sensors.

The hypothesis behind this work is that the context-aware continuous monitoring of the patient’s activities of daily living (ADL) will give a wider insight to doctors, thus allowing them to provide more appropriate feedback to patients and to more effectively design rehabilitation therapies. Moreover, such contextualized measures should open the door to the development of rehabilitation companions that provide feedback to the user for the sake of motivation and self- awareness.

In an increasingly older population and in the face of the current pandemic situation (Coronavirus disease), the central theme of this research, i.e., the continuous monitoring of patients in free-living conditions will certainly have a great impact and play a valuable role in our recent future.


Patients exhibiting motor control deficits, depending on severity, may encounter a range of difficulties in their daily life activities and they are prescribed to go through rehabilitation with the objective to optimize their daily life functional performance especially at home. While the actual performance of patients in the home environment is unknown Recent studies show that rehabilitation among patients with jagged movements can be enhanced when continuous monitoring of the quality of movement takes place during ADL.

With the aim of objectively monitoring the actual performance of patients in their home environment, we plan to evaluate their quality of movement. However, for measuring the quality of upper limb movements, in a systematic studies already mentioned that more than 150 different metrics have been used up to date. Some of those metrics measure the smoothness of the movement, others are based on other movement features like synergistic control and ergodicity, etc..

Enriching the objective measures obtained in unconstrained settings is feasible by taking advantage of context-awareness. Objectively assessing the movements’ quality in free-living based on just IMU sensors will provide a huge amount of context-free data. In order to be able to completely perceive and verify movements that are performed by patients at home, we need to employ a wearable camera. A system based on it has the potential to overcome the difficulties of continuous monitoring of patients at home.

Machine learning algorithms have already been used for recognizing the activities being performed by the user. For instance, activity recognition from videos is a widely studied challenge in the computer vision field. Although, analysis of videos captured from distant (third person views) grasped more attention so far than first person views.

All in all, having an intelligent system for context-aware monitoring of the patient’s daily life activities need to take advantage of machine learning techniques. The processing result of ego- centric videos through machine learning methods would give a wider insight to doctors, thus allowing them to provide more appropriate feedback to patients and further to more effectively design the rehabilitation therapies.

Study participants include adults more than 18 years old and younger than 65 years old with upper-limb neurological disorders who being inpatients at Nestlé Hospital du CHUV or outpatients followed either at Nestlé Hospital for rehabilitation. The aim is to cover different kinds of cerebral disorders. Participants will be medically stable and will be able to comprehend simple instructions.