The quantification of fundamental data about human posture and movement, such as balance and foot-ground interaction is a fundamental aspect for the evaluation of the quality of life of subjects with limited mobility, such as elderly persons [1
]. Most chronic disorders, like arthritis, diabetic foot, and neurodegenerative diseases result in a limitation of mobility and physical activity of the affected persons that, if not controlled on time, can rapidly compromise autonomy [2
]. Therefore, technological solutions that allow everyday monitoring of physical function can be used not only to follow the effects of medical conditions and treatments, but also to respond opportunely and prevent or slow down pathologic processes and risk conditions [3
One of the strategies to assess and monitor the physical condition in a non-intrusive way is by means of the analysis of the patient weight distribution against the floor using force platforms [4
]. However, since those instruments are usually bulky and expensive, their use is limited to specialized clinics and research labs and, therefore, cannot be used for long-term daily monitoring. Another possibility is the use of insole systems for monitoring plantar pressure during daily life activities [5
]. However, such an approach is usually limited by the low spatial resolution [6
] and implies wearing a device that may be difficult to wear or uncomfortable [7
Here we propose a new mat-like textile pressure mapping device that offers a reasonable trade-off between high-resolution, high-accuracy, but bulky and expensive, force platforms, and wearable, but low-resolution, insole systems. Our device is based on a pressure sensor made of a single layer of a 100% cotton fabric embroidered with silver coated polyester yarns and treated at specific spots with polystyrene sulfonate-doped poly-3,4-ethylenedioxythiophene (PEDOT:PSS). This results in a very thin and flexible textile device that can be easily rolled up during transport and storing and, thus, it can be quickly installed and removed from a home environment.
A similar solution may be developed with plastic film-based pressure sensors [8
], but a fiber-based alternative can offer superior comfort, flexibility, and mechanical robustness. Those properties allow being better adapted not only to frequent roll and unroll operations, but also to the integration with other textiles in the home environment (carpets, furniture covers, etc.
) and garments.
The idea of using textile technology for developing large area pressure sensors is not new [10
], but until now textile sensors either imply the use of multilayered configurations of fabrics and films that limit the flexibility and make intensive use of expensive materials [11
] or require the fabrication of specially-structured fibers [14
]. In contrast, the proposed sensor is based on conventional cotton fabric, allows using a relatively low amount of conductive polymers and metal-coated yarns, and relies on a fabrication process that deals with only one layer of fabric. Our sensor actually requires including two additional cover layers for providing moisture and mechanical protection, but such layers can be made of conventional low-cost fabrics and do not participate in the transduction process.
In this work, we present the development of a pressure sensitive mat prototype intended to be used in home environments for monitoring the physical condition of elderly persons by means of plantar pressure measurements. In particular, the mat is designed to obtain data about possible anomalies in foot-ground contact (which are typical of metabolic diseases like diabetes or musculoskeletal alteration, such as hallux valgus, toe deformities, flat foot, etc.
]) or assess static and dynamic balance abilities of an individual by analyzing center of pressure (COP) trajectories during static standing or sit-to-standing tasks.
The device is composed of 1024 sensing elements (i.e., sensels) arranged in a 32 × 32 array and covering an area of 39 cm × 39 cm which is enough to accommodate an average person standing with their feet shoulder breadth apart. The prototype includes multiplexing electronics for measuring pressure at high speed; programmable polarization and amplification circuits in order to compensate sensel-to-sensel variability; and modules for wireless LAN communication and web-based user interfaces in order to allow using it locally or remotelly through any LAN-connected device (e.g., SmartTV, tablet, etc.). A functional description of the sensor is presented together with a preliminary validation of the mat prototype for extracting plantar pressure parameters.
The resistive response of a single pressure-sensing element to a load of 5 N over 1000 cycles is shown in Figure 6
A. Though the zero load resistance shows some variability, the response to the applied load is stable. In Figure 6
B a typical loading-unloading curve is shown.
Single sensel and sensel-to-sensel variability of the mat prototype is illustrated in Figure 7
A,B, respectively. It can be noted that, though the response of a single sensel is reproducible, sensel-to-sensel variability can be quite high. This variability is due to a variety of reasons: for instance, the contact area between the upper and the lower yarns and the treated portion of the fabric is presumably not exactly the same for each sensor. Additionally, the amount of polymer adsorbed by the fabric depends on the fabric material and is not exactly reproducible in a manual deposition procedure. By automatically depositing the conductive compound, it should be possible to limit this source of variability, even if the textile structure of the fabric does not ensure a perfectly homogeneous density of the textile material in each sensor site. However, this variability is substantially reduced after applying the equalization procedure as observed from the resultant digital readout (Figure 8
A). Calibrated outputs of the same sensels are shown in Figure 8
and Figure 10
show examples of average plantar pressure maps from simultaneous recordings made with the proposed mat and a commercial platform during the bipedal and unipedal static trials, respectively (from different subjects). Similarly, Figure 11
show an example of maximum-pressure maps obtained during a walking (dynamic) trial. All maps were randomly selected from all recordings for illustrating a typical case for each type of trial. Pressure maps are scaled to illustrate real device dimensions and are compatible with existing methods of analysis. We did not apply any data transformation or sharpness enhancement algorithm.
Pressure values obtained with the mat show an overall agreement with those obtained with the platform even though a sensel-by-sensel inspection reveal some non-negligible differences. Table 1
allows making a quantitative comparison of this by showing the parameters calculated for each foot zone from the pressure maps shown on Figure 9
, Figure 10
and Figure 11
(only the right foot of Figure 9
was considered on Table 1
). It can be noted that that contact areas, mean pressures, and peak pressures calculated with the mat were numerically lower than the same parameters calculated with the platform, with the largest differences being observed in the midfoot, which is the region characterized by the lower pressure values.
gives a summary of relative percentage differences (RPD) obtained with all the static and dynamic trials performed. Data from bipedal and unipedal trials is shown in Figure 12
A,B respectively, whereas data from walking trials is shown on Figure 12
C. As already observed in the sample data of Table 1
, the RPD along all trials also show that differences are higher and more spread for midfoot and lower for rearfoot parameters. In general, the RPD for mean pressures are similar in static and dynamic conditions, whereas the RPD averages for peak pressures are lower for the dynamic (walking) measurements. The RPD averages for the contact area are much lower in dynamic measurements.
We have presented a textile pressure sensor with a very sensitive and stable resistive response. In contrast to capacitive textile sensors like those presented by Meyer et al.
] or by Lee et al.
], our sensor can be used with relatively simpler electronics at high-speed, is very thin (less than 1mm thick) and flexible, and uses a minimum amount of expensive materials. Such features are advantageous for large-area or wearable applications that require a high number of sensing elements while maintaining a low cost.
The main part of our sensor is fabricated by sewing and stamping a single layer of a conventional cotton fabric. While it still needs additional cover layers for providing moisture and mechanical protection, such layers are made of conventional low-cost materials and do not participate in the transduction process. Differently than what was reported by Shimojo et al.
], our approach, based on the point-like pressure sensors, avoids cross-talk effects and, in addition, is based on standard, easy-to-sew conductive yarns.
Our sensor is based on the conductive polymer PEDOT:PSS which has known piezoresistive properties that make it suitable for making force-sensitive sensors [27
]. We are aware, however, that it also has thermoelectric [29
] and humidity-sensitive [22
] properties. Preliminary experiments (see Supplementary Figure S1
) have confirmed that the impermeable cover effectively blocks the loss of conductivity associated with air moisture exposure. Although a precise characterization of the temperature response of our prototype has to be done, preliminary data suggest that temperature sensitivity of our sensor is negligible (see Supplementary Figure S2
) given that the target application here implies an indoor environment where temperature variations (presumably very few degrees celsius) arise mainly from the interactions with the human body through several layers of fabric.
The quantity and cost of the materials used in the proposed approach are relatively low when compared with existing solutions for measuring pressure distribution with flexible substrates [8
]. In terms of fabrication time and human effort, it took us just a few minutes to sew the conductive yarns with the automatic embroidering machine and the stamping of the conductive polymer spots, although done by hand by a single operator, took only two hours to complete. It is also possible to envision different automatic techniques for depositing the conductive polymer, such as screen printing, digital printing, etc.
Therefore, it is very likely that our approach also implies lower fabrication costs when compared with other textile or fiber-based pressure sensors [11
] if scaled to the same area as ours. Only the plastic film-based devices, like those presented by Tan et al.
] and Papakostas et al.
], may be fabricated with a faster and cheaper procedure since those are based on already industrialized sensors probably based on screen printing methods. However, achieving the same kind of performance on a fully textile platform, rather than on a plastic sheet, paves the way to a variety of new different applications, besides that shown here (i.e.
, the pressure sensitive mat). In fact, fiber-based materials offer superior comfort, flexibility and mechanical robustness [30
]. Those are properties that, for certain kinds of applications (for instance all those based on the contact with human body), are much more interesting than those based on plastics films [30
The possibility of using electronic embroidering machines allows us to control the features of our sensor with sub-millimeter resolution. The spatial resolution of our technology is, nevertheless, limited by the quality of the conductive yarn and not by the embroidering machines. This is because yarns are made of twisted fibers that can break and stick out during sewing, giving the yarn a hairy aspect. The “hairs” of the conductive yarns can be several millimeters long and produce short circuits between stitching lines that are supposed to be isolated. Given the 12 mm separation between the sensels of our mat prototype, we only observed such short-circuits in less than 2% of the sensels. By manually eliminating these hairs, we can easily go to a sensel separation as low as 7 mm; reasonably, this technology is not suitable for sensors with less than 5 mm of sensel separation.
Since both conductive yarns and substrate fabric are irregular at the microscopic scale, each sensel may absorb the conductive polymer solution in a different way and may have slightly different mechanical properties. Those may be the reasons why there is a high degree of variability in the resistive response between different sensels (see Figure 6
B). By carefully selecting a different type of fabric it might be possible to reduce this variability.
Another source of variability is given by the process of manually stamping the PEDOT:PSS on the cotton fabric which produces differences in the size of the treated area between individual sensels. However, since the surface resistivity of the treated fabric is much bigger than the resistivity of the conductive yarns (~10 kOhm/sq versus ~5 Ohm/cm), the current in a sensel flows mainly along the shortest path (i.e., the intersection point between the top and bottom conductive yarns). Therefore, it is plausible that variations of the treated area size affect much less the device performance than variations of the geometry and properties of the involved materials at the intersection itself. Most of the variations of the treated area can be easily reduced by automatizing the stamping process.
In any case, the equalization procedure allows us to reduce most of the variability (see Figure 8
A). Variability may be further improved by measuring the sensor response with instrumentation of higher quality.
The plantar pressure parameters calculated from the calibrated output of our mat prototype are in line with what has been observed for normal feet in studies reported in the literature [31
]. However, the data simultaneously acquired with the commercial platform show that most of the parameters are actually rather underestimated. We think that this is mainly a calibration problem that can be solved by both calculating a different set of calibration parameters for each sensel, and using a calibration setup that allows applying more accurate and homogeneous pressure values over the entire sensor. Since our textile sensor is not completely flat, the last is best achieved by using a setup entirely based on pressure (for instance, a pneumatic setup based on a caged rubber bladder) and not on the force distributed by a solid, like the one we used.
It is important to note that some of the discrepancies observed in the simultaneous measurements were due to the noise induced by the platform circuitry into the textile sensors. This is more evident on the maximum pressure map of Figure 11
(because it is the result of non-averaged data) where it is possible to observe some active sensels outside the actual footprint.
Although the proposed technology may need more development in order to achieve clinically-accurate plantar pressure measurements, results are encouraging and we believe that its applicability for home monitoring and fitness assistance is very promising. In our vision, a pressure mapping mat can be easily positioned in front of a couch or next to the bed of the user in order to monitor the plantar pressure and balance parameters during the sit-to-stand movement which is one of the indicators of the physical condition and falling risk in elderly people [32
]. Moreover, this kind of textile technology could be used in all those sensing tasks where flexibility and comfort are desired. For instance, mattresses and cushions for anti-decubitus applications could be another interesting application to explore.