近年來隨著人口高齡化，老年人口的增加為醫療單位帶來更大的負擔，因此如何透過醫療科技來降低住院意外的發生將成為醫院關注的重要議題，眾多住院意外當中又以跌倒對年長者的威脅最大，不但有立即之危險，留下的後遺症更會加重照護者與健保的負擔。為降低跌倒所造成的二度傷害，醫療院所多採用離床警報器於病人下床時主動通知護理師協助，然而目前市面上相關產品之假警報皆過於頻繁，不只導致照護者疲於奔命，在分秒必爭的醫院中更可能耽誤關鍵的急援時機。
基於上述議題，本團隊與成大醫院林宙晴主任團隊合作開發具低誤報率之新型離床警報系統。本論文將提出兩套新式離床警報器系統，其一為根據護理師臨床之照護經驗，於病人下床最常經過的路線上設置紅外線、超音波與三軸加速計等多重感測器，以監測病人有無離床之舉動，偵測結果會即時於護理站的電腦顯示。當病人離床時，除了護理站會接到警報提醒護理師盡快前往支援外，病床端亦會發出警報，以語音提示病人留候原地等候協助。本系統於醫院實測該系統對病人離床事件的陽性預測值可達85%，相較於傳統壓力式離床警報器增加了15%之精準度。
另一套離床警報器系統則使用裝置於床頭的紅外線陣列感測器捕捉病人之熱成像，再以支援向量機分析取得病人於床上的位置，判斷其是否要離床。不同於傳統影像處理，此法可省去邊緣偵測與動作追蹤等運算，並且不受環境光源影響。但支援向量機對單張影像之辨識率極低，因此本系統加入回授使該系統於驗證時能將病人先前的狀態納入考量，進而降低錯誤率至2%以下，且對病人離床事件之陽性預測值可達99%以上。

SUMMARY

This work presents two systems to prevent the elderly from fall. The first system uses infrared sensors, ultrasonic sensors and tri-axial accelerometers to sense the actions of the elderly. When the elderly wants to get out of bed, the system would stop him by recorded human voice and notify the caregiver to assist. Compared with similar systems on the market, the system reduces the false alarm rate to 15%. Another system uses an infrared array sensor to capture elders’ thermal images, and an SVM is used to analyze their positions on the bed. Error rate of the system is less than 2%, and the positive predictive value of patients' leaving events can reach more than 99%.

INTRODUCTION

In recent years, with population aging, increasing elderly population has brought a greater challenge to medical institutions. Therefore, it has become an important issue for hospitals to reduce the incidence of hospitalization accidents through certain technology. Among hospitalization accidents, falls are the greatest threat to the elderly, not only causing immediate injury, but also aggravating the burden of caregivers and National Health Insurance. In order to reduce the secondary injury caused by falls, medical institutions often use bed exit alarm devices to inform nurses to assist patients when they get out of bed. However, present products often send many false alarms, which make caregivers not only tired but also more likely to miss critical emergency time in hospitals.
In order to solve the above problem, our team developed a new bed exit alarm system with lower false alarm rates with the team of Dr. Chou-Ching K. Lin from National Cheng Kung University Hospital. Two new alarm systems are proposed in this paper. The first system uses infrared sensors, ultrasonic sensors and tri-axial accelerometers to sense the actions of the elderly. Based on the clinical experience of nurse practitioners, the sensors are placed on most common route of the patient leaving the bed to monitor whether the patient leaves the bed or not. And the detection results will be displayed immediately on the computer of the nursing station. When the elderly leaves the bed, the nursing station will receive an alert to remind the nurse to help him/her as soon as possible, and the bedside device will send a voice promptly to instruct the elderly to wait for assistance. Another system uses an infrared array sensor to capture elders’ thermal images, and an SVM is using to analyze their positions on the bed. Different from traditional image processing, this method can avoid edge detection and motion tracking, and is not affected by ambient light sources.

SYSTEM HARDWARE CONFIGURATION

The first proposed system consists of the following module: (1) Power supply module, (2) Micro-Controller Unit (MCU): ATMEGA2560, (3) Sonar module: SR-04, (4) Active IR Sensor, (5) Accelerometer & gyro module: MPU6050, (6) Wireless transmission module: DL-LN33, (7) Voice recorder module: ISD1820, (8) Class-D audio amplifier: PAM8403. The main MCU is a high-performance, low-power 8-bit microcontroller, it is used for processing the data from all module and sending information to computers. Sonar module, IR sensor and accelerometer are using to sense the elderly.
Another proposed system consists of the following module: (1) Power supply module, (2) Micro-Controller Unit (MCU): ATMEGA2560, (3) Infrared array sensor: AMG8833, (4) Wireless transmission module: DL-LN33. The infrared array sensor is mounted on the top side of the bed to get thermal images of the elderly.

SYSTEM FIRMWARE CONFIGURATION

There are two major firmware in the first proposed system. In the flowcharts of the first firmware, the main routine can be divided into two steps. In the first step, the system get readings from IR sensors and Accelerometer, and transmit them to the other firmware in the second step. In contrast, the main routine of the other firmware can be divided into three steps. The first step is receiving data from the previous firmware, and the second step is getting the readings of sonar modules to determine whether the elderly leaves the bed or not. The final step is sending all data to PC and show it to the caregivers.
In the flowcharts of the second system based on infrared array sensor, the main routine reads thermal images and ambient temperature and determine movements of the elderly using PC.

SYSTEM VERIFICATION AND RESULT

The first proposed system has been measured in the National Cheng Kung University hospital and has the 85% positive predictive value of the patient leaving the bed, which is better than the 70% positive predictive value using the traditional pressure bed exit alarm.
Another system based on infrared array sensor mounted at the head of a bed to capture thermal images of the elderly. At the beginning, the recognition rate of SVM for a single image is less than 20%, so feedback of the system is added to make the system consider the previous state of patients in verification. Finally, the error rate is reduced to be less than 2%, and the positive predictive value of patients' leaving events can reach more than 99%.

CONCLUSION

This work presents two bed exit alarm systems used to prevent the elderly from accident fall. For the system based on multiple sensors, the false alarm rate was reduced to half of the value using other products on the market. For the second system based on thermal imaging, infrared array sensors are used for detecting bed exit. After considering the continuity of human movement, the classification accuracy of patient status can reach more than 99%.

[1]. 內政部統計處─內政統計通報107年第六週
https://www.moi.gov.tw/stat/node.aspx?cate_sn=-1&belong_sn=7460&sn=7582
[2]. 衛福部病人安全通報系統2016年年報
http://www.patientsafety.mohw.gov.tw/Content/Downloads/List01.aspx?SiteID=1&MmmID=621273303702500244
[3]. A. Härmä, W. ten Kate, and J. Espina, “Bed Exit Prediction based on Movement and Posture Data,” IEEE-EMBS International Conference on Biomedical and Health Informatics, pp. 165-168, Jun. 2014.
[4]. R. L. S. Torres, D. C. Ranasinghe, Q. Shi, and A. P. Sample, “Sensor Enabled Wearable RFID Technology for Mitigating the Risk of Falls Near Beds,” IEEE International Conference on RFID, pp. 191-198, Jun. 2013.
[5]. K.-H.Wolf, K. Hetzer, H. M. Zu Schwabedissen, B. Wiese, and M. Marschollek, “Development and Pilot Study of a Bed-exit Alarm based on a Body-worn Accelerometer,” Zeitschrift für Gerontologie und Geriatrie, vol. 46, Issue. 8, pp. 727-733, Nov. 2013.
[6]. M. Pouliot, V. Joshi, J. Chauvin, R. Goubran, Fellow, IEEE, and F. Knoefel, “Differentiating Assisted and Unassisted Bed Exits using Ultrasonic Sensor,” IEEE International Instrumentation and Measurement Technology Conference, pp. 1-5, May. 2012.
[7]. D. Asbjørn and T. Jim, “Recognizing Bedside Events Using Thermal, and Ultrasonic Readings,” Sensors, vol. 17, no. 6, pp. 1342-1360, Jun. 2017.
[8]. T. Grant, V. Joshi, M. Taylor, F. Knoefel, H. Sveistrup, M. Bilodeau, and Jeffrey Jutai, “Measuring Sit-to-stand Timing Variability over Time using under Mattress Pressure Sensor Technology,” IEEE International Symposium on Medical Measurements and Applications, pp. 1-5, Jun. 2014.
[9]. H. Knight, J.-K. Lee, and H. Ma, “Chair Alarm for Patient Fall Prevention based on Gesture Recognition and Interactivity,” 30th Annual International IEEE EMBS Conference, pp. 3698-3701, Aug. 2008.
[10]. M. Bruyneel, W. Libert, and V. Ninane, “Detection of Bed-exit Events Using a New Wireless Bed Monitoring Assistance,” International Journal of Medical Informatics, vol. 80, no. 2, pp. 127-132, Feb. 2012.
[11]. T.-X. Chen, R.-S. Hsiao, C.-H. Kao, W.-H. Liao, and D.-B. Lin, “Bed-Exit Prediction based on Convolutional Neural Networks,” IEEE International Conference on Applied System Innovation, pp. 188-191, May. 2017.
[12]. B. Chwyl, A. G. Chung, M. J. Shafiee, Y. Fu, and A. Wong, “DeepPredict: A Deep Predictive Intelligence Platform for Patient Monitoring,” International Conference of the IEEE Engineering in Medicine and Biology Society Annual Conference, pp. 4309-4312, Jul. 2017.
[13]. Y. Li, L. Berkowitz, G. Noskin, and S. Mehrotra, “Detection of Patient’s Bed Statuses in 3D Using A Microsoft Kinect,” International Conference of the IEEE Engineering in Medicine and Biology Society Annual Conference, pp. 5900-5903, Aug. 2014.
[14]. P. Bauer, J. B. Kramer, B. Rush, and L. Sabalka, “Modeling Bed Exit Likelihood In A Camera-Based Automated Video Monitoring Application,” IEEE International Conference on Electro Information Technology, pp. 56-61, May. 2017.
[15]. N.M. Kosse, K. Brands , J.M. Bauer, T. Hortobagyi, and C.J.C. Lamoth, “Sensor Technologies Aiming at Fall Prevention in Institutionalized Old Adults: A Synthesis of Current Knowledge,” International Journal of Medical Informatics, vol. 82, no. 9, pp. 743-752, Sep. 2013.
[16]. P. Ward-Smith, L. Barrett, K. Rayson, and K. Govro, “Effectiveness of a Bed Alarm System to Predict Falls in an Acute Care Setting,” Clinical Nursing Studies, vol. 3, no. 1, pp. 1-4, Mar. 2015.
[17]. H.-M. Tzeng, and C.-Y. Yin, “Most and Least Helpful Aspects of Fall Prevention Education to Prevent Injurious Falls a Qualitative Study on Nurses’ Perspectives,” Journal of Clinical Nursing, pp. 2676-2680, Jan. 2013.
[18]. S. J. Cuttler, J. Barr-Walker, and L. Cuttler, “Reducing Medical-Surgical Inpatient Falls and Injuries with Videos, Icons and Alarms,” BMJ Journals, vol. 6, Issue. 2, pp. 1-9, Oct. 2017.
[19]. H. Liu, J. Huang, C. Lu, Z. Lan, and Q. Wang, “Indoor Monitoring System for Elderly Based on ZigBee Network,” International Symposium on Micro-NanoMechatronics and Human Science, pp. 1-7, Nov. 2016.
[20]. D.C. Ranasinghe, R.L. S. Torres, K. Hill, R. Visvanathan, “Low cost and Batteryless Sensor-Enabled Radio Frequency Identification Tag based Approaches to Identify Patient Bed Entry and Exit Posture Transitions,” Gait & Posture, vol. 39, no. 1, pp. 118-123, Jan. 2014.
[21]. A. Wickramasinghe, D. C. Ranasinghe, C. Fumeaux, K. D. Hill, and R. Visvanathan, “Sequence Learning with Passive RFID Sensors for Real-Time Bed-Egress Recognition in Older People,” IEEE Journal of Biomedical and Health Informatics, vol. 21, no. 4, pp. 917-929, Jul. 2017.
[22]. S. Mashiyama, J. Hong, and T. Ohtsuki, “A Fall Detection System Using Low Resolution Infrared Array Sensor,” IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, pp. 2109-2113, Sep. 2014.