"I need software that can tell me how many hours I slept, how long I was awake,
how much REM I had, core sleep, and deep sleep. Does your product provide this?"

We often have to do significant work to correct the misinformation generated by consumer gadgets, phone apps, rings, and watches, which claim to analyze sleep but are typically developed by engineers without a background in sleep science.

The metrics cited are useful in sleep studies, where data is collected from hundreds or thousands of people and it is impossible to examine each recording in detail.

For an individual, however, it is important to look at the actual sleep structure, much like you need to actually view an X-ray to assess whether a bone is healthy rather than relying on summary percentages of white, black and grey.

Sleep phenotype

The first thing you should be looking at is your sleep phenotype. Are you a frontal alpha producer (87% of people)? If so you will be able to detect alpha intrusion where present. This is an EEG anomaly not specifically addressed by common sleep metrics (in fact it can totally throw off the sleep staging in naive algorithms, which will mark epoch as Wake when the correct scoring is N2 or N3). Alpha producers can also rely on the alpha ridge on the spectrogram to better distinguish REM and Wake epochs.

Sleep fragmentation

Sleep can be fragmented. If you move in and out of slow-wave sleep every few epochs, your total sleep time may appear adequate, but the quality of sleep will be poor. Similarly, if you have difficulty falling asleep and experience frequent awakenings, it will no longer be possible to mark the first sleep cycle as having a clear beginning. In that case, any "sleep onset" metric is essentially meaningless, and it will distort the total sleep time calculation. This can vary night by night, so if you are attempting optimization, you need to see the actual sleep structure rather than relying on summary numbers.

Gaps and artifacts

It is also crucial to examine where signal gaps, artifacts, and electrode disconnections occur. If an electrode comes off during the night or another issue arises, those portions of data will be uninterpretable. For instance, if a headband is loose and you sleep with your head against the pillow, the breathing rhythm can appear on the EEG as slow frequencies. Simple consumer devices may interpret this as delta waves and report a falsely high total sleep index, whereas the spectrogram clearly shows a flood of frequencies that are immediately recognizable as non-sleep signals. Zooming in and scrolling through the signal makes this obvious.

It is orders of magnitude easier to develop frequency band filter based algorithms that distinguish the sleep phases in the absence of any artifacts (never the case) and for a typical phenotype (often not the case) - than to categorize, model and exclude the dozens of types of sleep artifacts that can occur on sleep EEG. For this reason most consumer devices that hide the raw data from the user will treat artifacts as EEG signal and produce highly unreliable results.

A brief list of artifacts that will affect sleep EEG analysis:

• Sweat
• Heart and breathing rhythm can be coupled to electrode impedance (ex loose headband pressed against the pillow)
• Eye movements: clearly discernible upon inspection, but often trick algorithms as their amplitude, slope and morphology differs from person to person and across REM segments
• Muscle activity (EMG): eye lid or jaw tension produces broad spectrum noise. On a spectrogram this is actually a useful phenomenon, as it immediately reveals arousals and awakenings, but throws off the frequency power calculations in naive algorithms
• Loose electrodes or poor contact (immediately visible on a spectrogram as dark or very bright vertical areas)
• Bruxism (immediately visible on a spectrogram as very bright vertical areas)

Breathing

Breathing is another aspect where examining the micro and meso structure of sleep is highly beneficial. Monitoring your breathing with the airflow/SpO2 sensor is strongly recommended, as adequate oxygenation is essential for restorative sleep.

Even a brief interruption in breathing of just a few seconds can disrupt sleep without being recorded as a notable abnormality by standard polysomnography. For example, a three-second pause may cause you to shift position and resume sleeping normally. Although it would not register as an apnea, it can still pull you out of slow-wave sleep entirely.

ZMax Total includes a sound pressure sensor to detect snoring and an airflow/SpO2 sensor that tracks airflow from both nostrils and blood oxygenation.

Sleeping position also plays a role, as one position may trigger apneas while another does not. This information can be directly actionable for improving sleep quality.

In summary

If your goal is to optimize sleep, reviewing the sleep structure is essential. The spectrogram alone provides most of the actionable information instantly, which is why we emphasize it so strongly.

The free software bundled with ZMax can generate a sleep report that includes metrics such as total time in REM, total slow-wave sleep, and deep sleep. You can see an example here.

However, we do not generally recommend relying on these metrics for individuals for the reasons outlined above. Furthermore, producing a report requires either some additional expense for our autoscorer, billed per night, or a small amount of manual work.

If you are interested in learning sleep scoring for normal (healthy) sleep EEG, tutorials are available on our YouTube channel. Becoming proficient at an amateur level can take as little as couple of hours, and the manual process takes roughly ten minutes per recording. This also allows you to understand what you are actually measuring. Once you create the hypnogram, the report is generated automatically.

For individual users, we generally do not recommend learning to score sleep. Instead,

• Focus on the spectrogram
• Review the shape of sleep onset, the structure of sleep cycles, and awakenings, which appear as vertical bars.
• You can also zoom into the signal to identify the cause of interruptions, such as sweat, noise, or breathing irregularities when using devices with airflow or SpO2 sensors.

Place an order on-line