0

Корзина пуста.

Оборудование в наличии

Quantitative Evaluation of Performance during Robot-assisted Treatment

Keywords

Assessment, robotics, rehabilitation, upper limb, cerebral palsy


Summary

Introduction:

This article is part of the Focus Theme of Methods of Information in Medicine on “Methodologies, Models and Algorithms for Patients Rehabilitation”.

Objectives:

The great potential of robots in extracting quantitative and meaningful data is not always exploited in clinical practice. The aim of the present work is to describe a simple parameter to assess the performance of subjects during upper limb robotic training exploiting data automatically recorded by the robot, with no additional effort for patients and clinicians.

Methods:

Fourteen children affected by cerebral palsy (CP) performed a training with Armeo®Spring. Each session was evaluated with P, a simple parameter that depends on the overall performance recorded, and median and interquartile values were computed to perform a group analysis.

Results:

Median (interquartile) values of P significantly increased from 0.27 (0.21) at T0 to 0.55 (0.27) at T1 . This improvement was functionally validated by a significant increase of the Melbourne Assessment of Unilateral Upper Limb Function.

Conclusions:

The parameter described here was able to show variations in performance over time and enabled a quantitative evaluation of motion abilities in a way that is reliable with respect to a well-known clinical scale.


Introducion

In clinical practice, it is substantial to idetify among the multiple variables involved in rehabilitation treatments which ones might have a larger impact on outcomes and influence recovery. Such evaluations require the use of quantifiable, valid, and sensitive tools to guarantee reliable between study comparisons and greatly improve the understanding of key treatment effects [1].

The main purposes of valid assessment methods have been largely discussed in the literature over the years. A useful test should be effective, efficient, comparable and predictable [2]. Unfortunately, many assessment methods commonly used today barely highlight the effectiveness of therapy treatments [3] since they are based on subjective impressions and could depend both on operators’ ability and on patients’ personality, and attitude. Further, they do not provide assessment during the training.

During the past few years, technologies gains an important role in rehabilitation [4] and robot-assisted rehabilitation has become a very active area of research [5, 6], since it provides controlled, intensive taskspecific training that is goal directed and cognitively engaging.

Measures derived from robots can contribute to the understanding of how different treatment variables (e.g., dosage, amount, and type of assistance provided) influence motor learning and recovery [7]. Between other advantages, robot-based data have higher resolution, better interrater and intra-rater reliability with respect to clinical scales and are able to quantify some physical quantities (i.e. kinematics and force data) that might be useful to investigate the neurorecovery. First attempts to extract meaningful information from these data are now present [3, 7, 8] and show that robot-based measurements can be correlated with clinical scales providing clinicians with data they need for assessing the patient’s capability, progresses during therapy, and ongoing therapy needs [3, 7].

Previous studies have developed some ad-hoc assessment tools to extract outcome measures of patients’ performance [8, 9], often requiring additional time for patients and clinicians. Others have exploited the built-it technology to extract indexes of task precision, movement smoothness and velocity [10]. Only few assessment methods were validated with functional scales [7, 11].

However the achievement of meaningful information from these data is still an ongoing challenge [3].

In this work we describe a simple pa rameter that can be easily derived from data saved by the robot and that quantifies the subjects’ performance. It can be used to follow the trend of a robot-aided treatment, to describe changes in performance before and after a rehabilitation and thus to investigate the effects of variations in the therapy on patients’ motor and functional recovery.


Materials and Methods

Fourteen inpatients (8 –16 y) affected by cerebral palsy (CP) performed a training with the paediatric Armeo®Spring (see Figure S1 in the supplementary online material), a five degree of freedom exoskeleton which guarantees passive arm weight support by means of springs, with a pressuresensitive handgrip and virtual reality visual feedback [12]. The research protocol was approved on March 2010 by the Ethics Committee of IRCCS E. Medea.


Experimental Design

Every subject underwent 15 –20 sessions lasting 30 minutes over 3 – 4 weeks of training depending on clinical suggestions. During each session, subjects performed a customized pull of exercises with the supervision of a physiotherapist. Eight exercises were selected to evaluate subjects’ performance over different joints and that require movements in different spaces (1D, 2D and 3D), accordingly to the indication of physiotherapists and clinicians. In particular, we evaluated one exercise performed in a 1D space (goalkeeper), five 2D exercises (egg cracking, fruit shopping, stove cleaning, moorhuhun and vertical catching) and two in a 3D space (chase balloon and reveal panorama). Details about the exercises are reported elsewhere [12]. During each training session, information about the exercise such as the scheduled difficulty level, the presence of the automatic grasping function, the thresholds to modulate the difficulty in grasping, the score obtained by the subject and the time required to perform the exercise were automatically recorded by the system, with no additional effort for the physiotherapists.

All these data have been included in a comprehensive performance parameter (pi) computed for the i-th exercise during each session, as in Equation 1:

where Si is the score obtained during the i-th exercise, Si,TOT is the maximum score obtainable, Ti is the time required to complete the ith exercise, Ti,TOT is the maximum time available to finish the i-th exercise and Di is the difficulty coefficient that considers the level of the i-th exercise and variation in autogrip and control threshold for each subject during the training.

In order to compare different exercises, pi has been divided over the maximum performance achieved in the i-th exercise by the group of subjects (Pi).

The median value of Pi over the eight selected exercises (P) was used to follow the training of each subject session by session. P is an index of the overall performance. Moreover, the median value of P within the first week (T0), between the 12th and the 16th days (T1/2) and within the fourth week (T1) of training was computed for each subject and finally the median value over the 14 subjects was calculated at the three time points.

Moreover we performed the Melbourne Assessment (M), which evaluates the level of impairment of upper limb motor function with a well-established inter-intrarater reliability [13], and we computed its the median value within the group at T0 and T1.

A validation of P has been proposed by comparing, for each subject, P variations with M variations referred to the maximum value obtained by the group.


Statistical Analysis

A non-parametric Friedman test for paired samples was performed on P values between T


Автор:  E. Peri / E. Biffi / C. Maghini / F. Servodio Iammarrone / C. Gagliardi