Ergonomic guidelines for the design interfaces of additive modules for manual wheelchairs: sagittal plane

When designing wheelchair propulsion systems operated with the upper limb, there is a noticeable lack of ergonomic analyses informing about the areas on the wheelchair frame where hand-operated controls can be installed. With that in mind, a research goal was set to measure the areas of human hand reach within the area defined by the structural elements of a manual wheelchair. An ergonomic analysis was performed on a group of ten patients representing 50% of anthropometric dimensions. Motion capture and image analysis software based on the openCV library were used for the measurement. The conducted research resulted in the development of a map of the hands range in the lateral plane of the wheelchair, parallel to the sagittal plane. In addition, the map was divided into three zones of hand reach, taking into account various levels of comfort of hand manipulation. The total hand reach area was 1269 mm long and 731 mm high, while the most comfortable manipulation area was 352 mm long and 649 mm high. The plotted hands reach areas act as a map informing the designer where on the sagittal plane additional accessories operated by the user can be installed.

www.nature.com/scientificreports/ physical features 16 . Currently, anthropometric atlases 17 are commonly used at the stage of design. However, there is a noticeable lack of criteria determining the comfort of using technical devices depending on the length of the kinematic chain of the human body used to operate it. The only available research on the comfort of use deals with the analysis of the influence of the seat position in relation to the push propulsion on the comfort of driving a wheelchair 18 . In addition, there is a noticeable lack of research linking the upper limbs reach with the positioning of the body in the wheelchair and the geometric features of the wheelchair frame. Among the available works, there are studies describing the impact of wheelchair configuration on its mobility 19 or studies analysing the type and frequency of activities performed in a wheelchair 20 .
In connection with the above considerations, a research goal was set. The objective was to measure the areas within the reach of the human hand within the structural elements of the manual wheelchair used to attach accessories that increase the functionality of the wheelchair. The premise for undertaking the research was the hypothesis assuming the designation of several zones in which additional wheelchair accessories can be attached. Each of these zones will use the kinematic chain of the human body of different lengths. The benefit of the obtained results is the preparation of guidelines for wheelchair designers, informing about where to place the control interface of additional modules, and in which zones the manual operation of the control interface should be simplified.

Method and materials
Research procedure. The research tests were carried out in real conditions of wheelchair use. A semi-active Vermeiren v300 wheelchair equipped with a hand position recording system was used for the study (Fig. 3). Hand range analysis was limited to the observation of the marker placed on the glove (Fig. 3d). The measuring system used consisted of a GoPro HERO 7 camera (Fig. 3a) and an illuminating lamp (Fig. 3b) mounted on the boom (Fig. 3c) permanently fixed to the frame of the wheelchair. The camera records the image in 960p quality with a speed of 240 fps. The illuminating lamp emitted 200 to 1000 lm, depending on the intensity of the ambient light. AruCo codes were used as markers 21 , printed on 50 × 50 mm plates.
The research procedure included the measurement of the position of the marker placed on the patient's hand (ID1) in relation to the stationary marker located in the wheelchair rear wheel axis of rotation (ID0) (Fig. 4). For the motion capture measurement, proprietary software was used that uses image processing of AruCo markers using OpenCV libraries that allow to determine their location in space relative to the immobile ID0 marker. It should be noted that other methods can be used to capture motion, e.g. using RGB-D Sensors 22 . Measurements were made in the sagittal plane 23 , which is parallel to the plane of the assumed XY datum. This approach is in line with the already developed two-dimensional wheelchair drive model 24 .   www.nature.com/scientificreports/ The research procedure includes the determination of four areas of hand manipulation with the use of motion capture measurement (Fig. 5). Each area was designated as the mean of the three measurement tests performed by each of the ten patients.
• The first area called the area of propulsion (AoP) defined the position of the hand while propelling the wheelchair (measured in real conditions during propelling the wheelchair). • The second area called area of comfort (AoC) defined the free movement of the kinematic chain consisting of the forearm (FA), while the arm (AR) was stationary and directed downwards parallel to the trunk (TR) (measured under quasi-static conditions). • The third area called the area of approval (AoA) defined the free movement of the kinematic chain consisting of the forearm (FA) and the arm (AR), while keeping the trunk (TR) stationary against the backrest of the wheelchair (measured under quasi-static conditions). • The fourth area called the area of risk (AoR) defined the free movement of the kinematic chain consisting of the forearm (FA), the arm (AR) and the trunk (TR), while keeping the hip motionless on the seat (measured under quasi-static conditions).
The segmentation of the human body used in the study complies with the accepted biomechanical standards 25,26 , and the adopted location of the ID1 marker on the hand complies with the anthropometric guidelines 16,27 . Tested men. Ten patients took part in the tests (Table 1) and they were subdivided by height, weight, age and wheelchair experience. Experience was determined on the basis of seniority in using the wheelchair. The patients participating in the tests reflected the 50th percentile of anthropometric dimensions according to the European standard "Basic list of definitions of human body dimensions for technical design" (EN 979). The decisive criterion for the patient's participation in the study was his anthropometric dimensions. Efforts were made to select the measurement group so that it represented the same percentile (50th percentile) of anthropometric dimensions. In addition, attention was paid to the similar age of the user and physical condition. This allowed  Data processing procedure. The data processing algorithm used to define the areas of hand manipulation by the manual wheelchair user in the sagittal plane involved five steps (Fig. 6). In the first step (Fig. 6a), the image recorded with the camera was converted into a set of points defined by two coordinates, horizontal "x" and vertical "y". For this purpose, proprietary software based on the OpenCV library was used 28 . The software recognised the position of marker ID1 relative to marker ID0. Then, using the alpha shape algorithm 29,30 , a designated closed-loop cloud was determined (Fig. 6b). When using the algorithm, the alpha coefficient ranged from 0.7 to 0.9, depending on the obtained point cloud density. Because after using the alpha shape algorithm,  www.nature.com/scientificreports/ each loop consisted of a different number of points, it was necessary to uniform their number. Therefore, using Rihnoceros 3D software, each loop was converted so that it would be described with a hundred of j P i points.
Where "j" is the index of the considered loop and "i" is the index of the considered point in the loop (Fig. 6c). Rhinoceros software with Grasshopper module was used for this activity. The next step was to search for all the designated loops of points between which the distance was the shortest. Using the points determined this way, the mean point P i , was calculated, which defined the averaged contour of the analysed area (Fig. 6d). The last stage was plotting the defined averaged areas on a common graph (Fig. 6e).

Results
At the beginning of the ergonomic analysis, characteristic geometric features were determined for the considered anthropotechnical system between the man and the wheelchair (Fig. 7) ( Table 2). The geometrical features defining the possible direction of extension and modification of the wheelchair are defined by the lower level of the frame (FDL), the upper level of the frame (FUL) and the level of the armrests (ARL). The levels defining the position of the human body in the wheelchair are described by the seat level (SL), the seat backrest vertical angle (SBV), the rear wheel spinning area (WB) and the shoulder height (SH). In this comparison, the shoulder level is the only anthropometric dimension determined experimentally, therefore the confidence interval (SH MIN and SH MAX ) was determined for the measurement sample N = 10 and the confidence level p = 0.05. The geometrical features SBV and SL define the position of the seat in relation to the axis of rotation of the drive wheel. They are often selected individually by each user 1,31 . Therefore, in the adopted ergonomic model, they can be freely modified. However, when modifying the features (in relation to the data given in Table 2), the designated areas of reach of the hand of a person sitting in a wheelchair should be properly shifted vertically and horizontally. These shifts should be the same as the changes in the seat position in relation to the data adopted in the work. To determine the geometrical features FDL, FUL and ARL defining the dimensions of the frame of the wheelchair, the dimensions of the frame of a semi-active wheelchair popular in Europe were used. These features can also be freely changed depending on the adopted type of wheelchair.
Having defined permanent geometrical features of the man and the wheelchair, the areas of reach of the hand of the person sitting in a wheelchair were determined (Fig. 8). These ranges were determined by subdividing them into four areas. The designated areas are the average value of 30 tests (N = 30) for which the confidence interval was determined for the confidence level of 95% (p = 0.05). The additional materials (Supplementary Table 1-4, Supplementary Fig. 1-4) contain a complete set of 100 points describing each area. Additionally, four control points were determined for each of the areas ( Table 3). The points illustrate the height and width of each field and serve as control points that can be used to shift areas in the sagittal plane when analysing wheelchair types other than the ones used in the tests.
To interpret the obtained results, the lengths of the confidence intervals were calculated for the adopted control points Δ (Eq. 1). www.nature.com/scientificreports/ When analysing the results, it was found that AoR is the largest area of hand manipulation. This area is 1269 mm long and 731 mm high. However, the greatest values of the confidence interval Δ were measured for the area located between extreme values of control points. These distances were on average 258 mm. For comparison, the average distance Δvalues for the AoA and AoC areas were 140 mm and 98 mm, respectively. The phenomenon of such differences in the lengths of the confidence intervals in individual areas is the effect of a different number of kinematic chains body segments 32,33 used during their determination. In the case of the AoR area, significant differences between the mean value of the area and its dimensions taking into account the confidence interval result from the share of the trunk in the kinematic chain used. Each tested patient had different mobility of trunk, so the greatest differences in hand reach were examined 34 . Consequently, the AoR has been classified as ergonomically risky to install the accessory interface of the manual wheelchair. The ergonomic risk of installing the interface in this area results from the inability to clearly determine the limit inclinations of the trunk at which the user is not at risk of falling from a wheelchair. Differences in the trunk angle between individual patients result not only from physical abilities 35 but also from wheelchair accessories such as seat belts 36 .
The AoA area in which the trunk was resting on the seat was 1009 mm in length and 553 mm in height, so the dimensions slightly smaller than those in the AoR area. However, in this case, much shorter lengths Δ were noted between the extreme positions of the control points and the control points with consideration of the confidence interval. This translates into greater precision in defining the reach of the hand. Additionally, a person who is physically disabled using a kinematic chain consisting only of the upper arm and forearm has the trunk stabilised. As a result, the risk of falling when tilting the trunk is minimised 37 .
The AoC area has been defined as the most comfortable one because it involves the shortest kinematic chain including the forearm only. As a result, the person using the wheelchair is subjected to the least physical effort 38 .   www.nature.com/scientificreports/ Despite these advantages, this area had the smallest surface area. Its length was 352 mm and the height was 649 mm. The last AoP examined described the area of hand manipulation while propelling the wheelchair. This area defines the place that should be free from additional accessories and should allow the user to freely manipulate it while propelling the wheelchair. The smallest length Δ dispersion was observed for the AoP area between the extreme positions of the control points and the control points with consideration of the confidence interval. The length of the AoP area was 412 mm and its height was 245 mm. It should be noted that the designated area of AoP coincides with the trajectories of movement of the hand pushing the handrim [38][39][40] .

Discussion
According to the methodological assumption, the defined permanent geometrical features of the human body and the wheelchair, as well as the measured areas of reach of the hand were superimposed. As a result, three areas of reach of the hand were obtained within the lower edge of the frame and the armrest of the wheelchair (Fig. 9). In order to implement the results in design problems, these areas were defined using a set of points enabling their approximate plotting ( Table 4).
The designated areas illustrates the three comfort zones for hand manipulation, located within three levels defined by the geometric features of the wheelchair. The area between the FUL and FDL levels illustrate the places on the wheelchair frame that do not change their position in relation to the drive wheel. The area between the ARL and FUL levels is where the armrests are located. This area was separated because during the use of the wheelchair (e.g. for dismounting 41 ) they are disassembled. Therefore, when planning to install additional accessories on these elements, one should take into account the need for its occasional disassembly. The performed subtraction of the previously determined areas showed that the most comfortable AoC in terms of safety and effort reduction is significantly reduced by the area reserved for the spinning wheel WB (Fig. 7) and the area in which the hand that pushes the handrim (AoP). After considering these areas, the AoC area is 194 mm wide and 227 mm high. www.nature.com/scientificreports/ The main determinant of the shape of the developed areas ( Fig. 9) is the length of the trunk, arm and forearm. Comparing the obtained results, a similarity with the data available in anthropometric atlases was noticed. An example of this is point 2 ( Table 4, Fig. 9) in which the torso was leaning against the backrest (SBV) and the upper limb was straightened and in a position close to horizontal. For such a system, the value on the x-axis representing the length of the arm and forearm is 753 mm, this value coincides with the data from anthropometric atlases. According to which the length of the upper limb measured from the axis of rotation of the shoulder joint to the center of the clenched fist is 783 mm 41 , 765 mm 42 and 743 mm 43 for a 50th percentile male. The same relationship Table 3. List of control points in the designated areas of hand reach. Where: x and y location of the control point on the sagittal plane, δx confidence interval of the control point location on the horizontal axis, δy confidence interval of the control point location on the vertical axis, Δ the distance between the extreme positions of the control point, n the size of the sample used to calculate the control point position, p adopted confidence level.

Point ID
x (mm) y (mm) δx (mm) δy (mm) Δ (mm) n (n/a) p (n/a)  Referring to the vertical and horizontal limitations marked in Fig. 9: FDL, FUL, ARL, SL, SH and SVB and their location in space, it should be noted that these are constant values that do not depend on the position of the human body, only on the design features of the wheelchair. They result from the geometric features of the wheelchair structure used. These are shaved available data that can be downloaded from the website of the manufacturer of the tested wheelchair. In addition, these data are available in the instruction manual supplied with the wheelchair. This approach makes the results valid for the cross-frame semi-active wheelchair group. This introduces a certain limitation and inconsistency of the map developed in Fig. 9 with wheelchairs from the group of active wheelchairs with a fixed frame. However, for these wheelchairs, it is possible to implement designated hand-reach areas that do not take into account the design features of the wheelchair (Fig. 8, Table 3).

Conclusions
The plotted AoC, AoA and AoR areas (Fig. 9) can be used as a map informing the designer where on the sagittal plane additional accessories operated by the upper limb of the user can be installed. The authors proposed a subdivision into three areas, depending on the range of the kinematic chain used. The AoC area was classified as the most comfortable because the movement of the hand within it translates into the least effort. The AoR area illustrated the limits of hand ranges of a man sitting in a wheelchair. The area was classified as the least comfortable because manipulation within its limits may increase the effort and the risk of falling off the wheelchair. The classification of areas presented in the work confirms the research hypothesis about the possibility of dividing the reach of the hand into different areas where the operation of the interface requires different lengths of the kinematic chain.
Using the developed ergonomic analysis, the designer can design additional accessories on the frame of the wheelchair. The accessories should be arranged according to the frequency of operation and the complexity of manipulations performed by the hand. Separated areas are helpful when arranging accessories. The accessories used frequently or requiring the use of increased physical force should be located within the AoC or AoA areas. Whereas the accessories that are used sporadically or do not require the use of significant force may be located within the AoR area.
When determining the AoR and AoE area contours, patients were not asked to raise the forearm vertically upward because the determined hand reach areas extend far beyond the designated levels of wheelchair geometric features. Such limitation of the entire limb range of motion did not affect the research goal and allowed us to bring the focal length of the camera closer to the tested object. Zooming in the focal length increased the accuracy of the measurement.
The conducted research supplements the deficiencies in anthropometric measurements, most of which focus on the reach of the hand of a person that is physically fit. In addition, the ergonomic analyses currently available are of generic nature and require further processing for implementation in the wheelchair design process. The developed map of hand movement ranges during manipulation within the wheelchair frame is equivalent of the available analyses of the geometry of the human workplace 45 that currently the basis for the development of standards certifying offices, schools, etc.
The results presented in the study show only the areas lying on the sagittal plane. This plane was chosen due to the fact that most of the additional accessories, such as the parking brake, are mounted on this plane. In further works, the authors plan to examine the hand reach area in the front and rear planes of the wheelchair, parallel to the head plane.  www.nature.com/scientificreports/ When subdividing the hand reach areas, a classification based only on the length of the kinematic chain was used. Therefore, in further ergonomic analyses, it is possible to link the position of the hand to the measurement of the EMG signal measuring muscle activity. This will allow for a new subdivision of the hand reach into areas according to the different muscle activities.

Data availability
The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.