利用室內空氣資料進行呼吸道疾病預測模型
No Thumbnail Available
Date
2022
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
在高齡化時代，台灣約有1.5%的老人人口安置於長期照護機構內。隨著年齡增長，長者往往有較脆弱的免疫系統以及肺功能，因而也有較高的危險性罹患呼吸道相關的疾病。由於長照機構通常為較密閉的空間，因此室內通風與空氣品質較為不佳。當有長者感染呼吸道疾病，造成群聚感染的風險也相對較高，而同時呼吸道疾病也會加重長者衍生成合併症及致死程度的危險性。如果能在長者感染的初期就提出警訊，以及早警告醫護人員並對其進行診斷及隔離等措施，就能夠降低長照機構中群聚感染發生的機率。本研究提出了運用自動編碼器的無監督式異常檢測室內空氣品質方法 (Unsupervised Anomaly Detection using Indoor Air Quality with Autoencoder for Respiratory Disease Prevention, AutoUAD-IAQ)以預防呼吸系統疾病傳染。初期本研究利用微型空氣感測器－Airbox蒐集長照機構的室內空氣品質，同時記錄長照機構感染呼吸道疾病的人數與時間。結合蒐集的資料並利用提出的呼吸道疾病預測模型方法來預測罹患呼吸道疾病的可能性，並在感染初期提出警訊以預防群聚感染的發生。
In Taiwan, approximately 1.5% of the elderly population is placed long-term care facilities (LTCFs). With aging, older residents and patients in LTCF tend to have weaker immune systems and lung function. Thus, this put them at higher risk of infecting respiratory diseases. Since most of LTCFs are in confined spaces, indoor ventilation and air quality are usually below the normal condition. When there is a respiratory infected patient in LTCF, the risk of cluster infection is high which can quickly spread and cause the risk of complications or fatalities for other residents or patients. If warning signs can be detected during the early stages of respiratory disease infection, then medical staff can diagnose and isolate the infected patient to reduce the probability of cluster infection in LTCFs.In this research, we proposed an Unsupervised Anomaly Detection using Indoor Air Quality with Autoencoder for Respiratory Disease Prevention (AutoUAD- IAQ) to prevent respiratory disease infection. First, a number of sensor devices, called AirBox, are deployed in residents’ rooms to collect indoor air quality data in LTCFs. Also, the medical records of respiratory diseases, such as the number of cases and duration of each case, are provided by individual LTCFs. Combining the collected data with the proposed AutoUAD-IAQ to predict the possibility of developing respiratory diseases, alerts can be raised at the early stage of infection to prevent the occurrence of cluster infections or outbreaks of respiratory disease.To verify the relationships between indoor air quality and respiratory disease, a cross-over study with Logistic regression was conducted using the collected data in LTCFs. The result confirms that temperature, humidity, and PM particles have a high degree of correlation with the risk of infecting respiratory diseases. Once the correlation is confirmed, machine learning and artificial neural network are used to forecast respiratory disease. The result shows that AutoUAD-IAQ has a 99% of accuracy in predicting respiratory diseases. With the proposed model, an early warning can be provided to LTCF’s staff can take appropriate actions to prevent cluster infections or outbreaks of respiratory disease in the early stage.
In Taiwan, approximately 1.5% of the elderly population is placed long-term care facilities (LTCFs). With aging, older residents and patients in LTCF tend to have weaker immune systems and lung function. Thus, this put them at higher risk of infecting respiratory diseases. Since most of LTCFs are in confined spaces, indoor ventilation and air quality are usually below the normal condition. When there is a respiratory infected patient in LTCF, the risk of cluster infection is high which can quickly spread and cause the risk of complications or fatalities for other residents or patients. If warning signs can be detected during the early stages of respiratory disease infection, then medical staff can diagnose and isolate the infected patient to reduce the probability of cluster infection in LTCFs.In this research, we proposed an Unsupervised Anomaly Detection using Indoor Air Quality with Autoencoder for Respiratory Disease Prevention (AutoUAD- IAQ) to prevent respiratory disease infection. First, a number of sensor devices, called AirBox, are deployed in residents’ rooms to collect indoor air quality data in LTCFs. Also, the medical records of respiratory diseases, such as the number of cases and duration of each case, are provided by individual LTCFs. Combining the collected data with the proposed AutoUAD-IAQ to predict the possibility of developing respiratory diseases, alerts can be raised at the early stage of infection to prevent the occurrence of cluster infections or outbreaks of respiratory disease.To verify the relationships between indoor air quality and respiratory disease, a cross-over study with Logistic regression was conducted using the collected data in LTCFs. The result confirms that temperature, humidity, and PM particles have a high degree of correlation with the risk of infecting respiratory diseases. Once the correlation is confirmed, machine learning and artificial neural network are used to forecast respiratory disease. The result shows that AutoUAD-IAQ has a 99% of accuracy in predicting respiratory diseases. With the proposed model, an early warning can be provided to LTCF’s staff can take appropriate actions to prevent cluster infections or outbreaks of respiratory disease in the early stage.
Description
Keywords
none, none