Macey P M, Ford R P, Brown P J, Larkin J, Fright W R, Garden K L
Department of Electrical and Electronic Engineering, University of Canterbury, Christchurch, New Zealand.
Acta Paediatr. 1995 Oct;84(10):1103-7. doi: 10.1111/j.1651-2227.1995.tb13505.x.
We examined the consistency of apnoea recognition between three human experts. The hypothesis was that computer detection of apnoea could emulate human expert apnoea recognition. The aim was to detect apnoeas with the highest possible accuracy from a single breathing signal, by both human experts and computer. Three human experts independently examined recordings of breathing wave-form from overnight sleep studies from 10 infants aged 3-17 weeks. All apnoeas of 5 s or more were identified and reviewed. However, there still remained 10% disagreement. A computer apnoea detector was implemented. An algorithm analysed statistical properties of the signal to find breathing pauses. Optimal performance was 1% missed apnoeas (compared with the agreed apnoeas identified by the three experts) and 29% false detections. This computer algorithm reliably identified most apnoeas but did not replace the human expert.
我们检测了三位医学专家对呼吸暂停识别的一致性。假设是计算机对呼吸暂停的检测能够模拟医学专家对呼吸暂停的识别。目的是由医学专家和计算机从单一呼吸信号中尽可能准确地检测出呼吸暂停。三位医学专家独立检查了10名年龄在3至17周的婴儿夜间睡眠研究中的呼吸波形记录。所有持续5秒或更长时间的呼吸暂停均被识别并复查。然而，仍存在10%的分歧。实施了一种计算机呼吸暂停检测器。一种算法分析信号的统计特性以找出呼吸暂停。最佳性能为漏检率1%（与三位专家共同认定的呼吸暂停相比），误检率29%。这种计算机算法能可靠地识别出大多数呼吸暂停，但无法取代医学专家。
