Skin Sensing Research Group, School of Health Sciences, Faculty of Environmental and Life Sciences, University of Southampton, Southampton SO17 1BJ, UK.
School of Engineering, Faculty of Engineering and Physical Sciences, University of Southampton, Southampton SO17 1BJ, UK.
Sensors (Basel). 2023 Aug 2;23(15):6872. doi: 10.3390/s23156872.
Commercial pressure monitoring systems have been developed to assess conditions at the interface between mattress/cushions of individuals at risk of developing pressure ulcers. Recently, they have been used as a surrogate for prolonged posture and mobility monitoring. However, these systems typically consist of high-resolution sensing arrays, sampling data at more than 1 Hz. This inevitably results in large volumes of data, much of which may be redundant. Our study aimed at evaluating the optimal number of sensors and acquisition frequency that accurately predict posture and mobility during lying. A continuous pressure monitor (ForeSitePT, Xsensor, Calgary, Canada), with 5664 sensors sampling at 1 Hz, was used to assess the interface pressures of healthy volunteers who performed lying postures on two different mattresses (foam and air designs). These data were down sampled in the spatial and temporal domains. For each configuration, pressure parameters were estimated and the area under the Receiver Operating Characteristic curve (AUC) was used to determine their ability in discriminating postural change events. Convolutional Neural Network (CNN) was employed to predict static postures. There was a non-linear decline in AUC values for both spatial and temporal down sampling. Results showed a reduction of the AUC for acquisition frequencies lower than 0.3 Hz. For some parameters, e.g., pressure gradient, the lower the sensors number the higher the AUC. Posture prediction showed a similar accuracy of 63-71% and 84-87% when compared to the commercial configuration, on the foam and air mattress, respectively. This study revealed that accurate detection of posture and mobility events can be achieved with a relatively low number of sensors and sampling frequency.
商业压力监测系统已被开发出来，用于评估处于发生压疮风险的个体的床垫/坐垫界面的状况。最近，它们已被用作长时间姿势和活动监测的替代方法。然而，这些系统通常由高分辨率的传感阵列组成，以超过 1Hz 的频率采样数据。这不可避免地导致了大量的数据，其中大部分可能是冗余的。我们的研究旨在评估准确预测躺着时姿势和活动的最佳传感器数量和采集频率。使用连续压力监测器（ForeSitePT，Xsensor，卡尔加里，加拿大），带有 5664 个以 1Hz 采样的传感器，评估在两种不同床垫（泡沫和空气设计）上进行躺着姿势的健康志愿者的界面压力。这些数据在空间和时间域中进行了下采样。对于每种配置，估计了压力参数，并使用接收者操作特征曲线下的面积（AUC）来确定它们区分姿势变化事件的能力。卷积神经网络（CNN）用于预测静态姿势。在空间和时间下采样时，AUC 值呈非线性下降。结果表明，采集频率低于 0.3Hz 时，AUC 会降低。对于某些参数，例如压力梯度，传感器数量越低，AUC 越高。与商业配置相比，在泡沫和空气床垫上，姿势预测的准确率分别为 63-71%和 84-87%。这项研究表明，通过相对较少的传感器数量和采样频率，可以实现对姿势和活动事件的准确检测。
