斯利普兰:信任始于对自动睡眠分阶段模型的公平评估
Alvise Dei Rossi1,2, Matteo Metaldi1, Michal Bechny1,3
1Institute of Digital Technologies for Personalized Healthcare ∣ MeDiTech, University of Applied Sciences and Arts of Southern Switzerland, Lugano, Switzerland.
NPJ digital medicine
|December 16, 2025
概括
我们开发了SLEEPYLAND,这是一个拥有大量睡眠数据的开源框架,以及SOMNUS,这是一个集成模型,可以显著提高不同数据集的睡眠阶段准确性和概括性. 在某些情况下,SOMNUS的表现优于个别模型,甚至超过了人类得分者.
科学领域:
- 人工智能的人工智能
- 睡眠医学 睡眠医学
- 生物医学工程 生物医学工程
背景情况:
- 对于自动睡眠分阶段的深度学习显示出希望,但在概括,偏见和评估方面面临挑战.
- 由于这些问题,临床采用是有限的,这阻碍了AI在睡眠分析中的广泛使用.
研究的目的:
- 推出SLEEPYLAND,一个全面的开源框架,拥有大量的睡眠阶段研究数据集.
- 开发和评估SOMNUS,这是一个设计用于强大和通用睡眠阶段的合奏模型.
- 评估模型概括,偏见和与人类得分器相比的性能.
主要方法:
- 利用了来自不同群体和硬件的~220,000小时的域内数据和~84,000小时的域外多睡眠数据.
- 开发了SOMNUS,一个使用软投票集体模型,以集成多个预训练的深度学习模型.
- 评估了24个数据集的模型,包括单通道和多通道EEG/EOG,并使用Bern-Sleep-Wake-Registry分析了人口/临床偏差.
主要成果:
- 在24个数据集 (宏观F1,68.7-87.2%) 中,SOMNUS实现了强大的性能,在94.9%的案例中表现优于单个模型.
- 索姆纳斯展示了改进的概括性,但并没有在所有架构中始终尽量减少人口/临床偏差.
- 在多注释数据集中,SOMNUS超越了最佳人类得分者 (宏观F1,85.2%与DOD-H的80.8%相比),集团分歧指标预测了得分者模糊性 (ROC-AUC 82.8%).
结论:
- SLEEPYLAND和SOMNUS提供了一个有价值的资源和一个高性能,通用解决方案,用于自动睡眠阶段.
- 虽然整体方法提高了稳定性,但解决模型偏差仍然是人工智能睡眠阶段化未来研究的关键领域.
- 整体分歧指标提供了一个有希望的方法来量化和预测人类得分者在睡眠阶段化任务中的不确定性.
相关概念视频
Understanding Sleep
1.4K
Sleep, an essential biological state, involves significant reductions in physical activity, sensory awareness, and interaction with the environment. This complex physiological process is primarily regulated by specific brain regions, notably the hypothalamus and pons, which govern the sleep-wake cycle or circadian rhythm.
The circadian rhythm, a nearly 24-hour cycle, is deeply influenced by environmental light cues. Light exposure directly affects the hypothalamus, which in turn regulates...
The circadian rhythm, a nearly 24-hour cycle, is deeply influenced by environmental light cues. Light exposure directly affects the hypothalamus, which in turn regulates...
1.4K
Stages of Sleep
1.3K
Sleep progresses through distinct stages, each characterized by specific brain wave patterns and physiological responses ranging from wakefulness to stages of non-rapid eye movement, known as non-REM, to rapid eye movement, referred to as REM. Understanding these stages helps in recognizing how sleep supports various bodily and cognitive functions.
Before sleep begins, in wakefulness, the brain exhibits primarily beta waves, which are high in frequency and low in amplitude, indicating alertness...
Before sleep begins, in wakefulness, the brain exhibits primarily beta waves, which are high in frequency and low in amplitude, indicating alertness...
1.3K
Blind Procedures
12.8K
Ideally, the people who observe and record the children’s behavior are unaware of who was assigned to the experimental or control group, in order to control for experimenter bias. Experimenter bias refers to the possibility that a researcher’s expectations might skew the results of the study. Remember, conducting an experiment requires a lot of planning, and the people involved in the research project have a vested interest in supporting their hypotheses. If the observers knew which...
12.8K
Optimal Arousal Theory
680
The optimal arousal theory suggests that performance is maximized when an individual experiences a moderate level of arousal. This theory is closely tied to the Yerkes-Dodson law, which illustrates an inverted U-shaped relationship between arousal and performance. The law, formulated by psychologists Robert Yerkes and John Dodson, implies an ideal arousal level for optimal performance, and deviations from this level can lead to declines in effectiveness.
Inverted U-Shaped Performance Curve
The...
Inverted U-Shaped Performance Curve
The...
680
Sleep-Wake Cycles
2.7K
Sleep is an essential physiological process vital to maintaining overall well-being. The reticular activating system (RAS), a network of neurons in the brainstem, regulates wakefulness and sleep. While it may seem passive, sleep consists of distinct cycles, each with its unique characteristics and functions. Two key sleep phases are non-rapid eye movement (NREM) and rapid eye movement (REM).
NREM Sleep
NREM sleep comprises four progressive stages that seamlessly merge:
NREM Sleep
NREM sleep comprises four progressive stages that seamlessly merge:
2.7K


