If you snore, you are probably familiar with the following symptoms: nights of poor sleep,, headaches, fatigue, daytime drowsiness, and constant complaints from the person sleeping beside you. At first glance, it may seem normal and unavoidable, right? However, many people do not realize that snoring might be just the “tip of the iceberg”, a symptom that could signal a more serious underlying condition.
One potential cause of loud and frequent snoring is a disorder called obstructive sleep apnea (OSA). This condition is characterized by partial or complete blockage of the airways during sleep, making it difficult for air to pass through.
Other contributing factors include obesity, smoking , and excessive alcohol consumption.Recent studies have even identified a genetic predisposition to snoring, some genes can increase the likelihood of developing this problem. So, if you have relatives who snore or if any of these risk factors sound familiar, stay alert!
So, What Is Snoring?
Snoring is the sound produced by the vibration of structures in the upper airway (such as the soft palate, tongue, tonsils, and epiglottis) during breathing. Imagine a crystal glass: when gently tapped with a spoon, it emits a sharp, distinct sound. Something similar happens in the body when we snore, though the result is far less elegant. Snoring is typically inconsistent, varying in volume, frequency, and duration.
What is known as regular snoring that is, the kind associated with sleep apnea, is very common within the population. It affects between 35% and 45% of men and 15% to 28% of women. It is also known that age influences sleep apnea. Nowadays, it is estimated that half of the people over 60 snore.
This does not sound like a great scenario, does it? The good news is that there are multiple treatment options available, depending on the cause. These include lifestyle changes, and weight loss,the use of oral appliances, and positional therapy(such as sleeping on your side). If snoring is linked to sleep apnea, Continuous Positive Airway Pressure (CPAP) (therapy is often recommended. CPAP devices act like silent air compressors to keep airways open. In some cases, healthcare practitioners may also consider surgery.
How Can You Find Out If You Snore, Have Sleep Apnea or Are Responding Well to Treatment?
It is actually quite simple! Using a compact sensor and a mobile app, you can perform the Biologix Sleep Test , , which analyzes oximetry data (that is, blood oxygen levels) to calculate the Oxygen Desaturation Index (ODI). In addition, it provides valuable information on heart rate, hypoxic burden, and movement.
The Biologix solution also offers an important extra feature: it detects and classifies the patient’s snoring, indicating both its intensity and the percentage of time spent snoring during sleep. This data is highly useful for diagnosing sleep disorders and monitoring patient progress throughout different therapies. Based on the collected data, the test result is classified as normal, mild, moderate, or severe apnea.
How Does the Biologix Solution Classify Snoring?
To accurately detect snoring patterns, the Biologix solution uses artificial neural networks, an automatic learning technology known as machine learning . This technology enables Biologix’s snoring-detection algorithms to learn continuously and adapt to new scenarios, mimicking some of the cognitive processes humans use to solve problems. You might think this concept sounds unfamiliar, but it is already part of your daily life; your streaming platform’s personalized recommendations, for example, rely on machine learning.
Machine learning works much like teaching a child, adult, or even a pet: we describe characteristics and provide examples. Imagine you’ve never seen an apple or a banana, we might explain that apples are typically round, red or green, with smooth skin, while bananas are long, yellow, and peelable. After seeing several examples, you would start to recognize them easily. The same process applies to machines.
In the case of Biologix, we define snoring characteristics (such as a deep, rhythmic sound lasting between 0.6 and 2 seconds) and feed multiple examples into the neural network. For this training, audio recordings from 268 patients were used, all collected during polysomnography exams at the InCor Sleep Laboratory at the University of São Paulo Medical School. Each audio file was reviewed by a human evaluator who manually labeled the sounds as “snoring” or “noise.”
Data Input
Raw data is recorded for classification, whether it's text, images, or, like in this case, audio.
Numerical Conversion
Data is converted into numbers. For example, an image is turned into a matrix where each number corresponds to a pixel’s color.
Feature
Extraction
Key characteristics are identified. For snoring, this includes volume, frequency, and duration.
Learning
The neural network calculates the impact of different values for each characteristic. Those that contribute most to accurate classification receive higher scores.
Classification
The network classifies sounds based on the information it has learned during training.
Next, we will list the previously presented examples, along with the characteristics that each neural network identified as relevant for initiating its training and learning process. In this methodology, each attribute is assigned a score, with the most influential ones receiving the highest scores. In the following stage, the network is presented with new samples, allowing it to apply the acquired knowledge by accurately classifying the sound events. Once this process is completed, the performance of the tool is evaluated using various statistical criteria, such as accuracy, sensitivity and specificity. The better these metrics, the greater the confidence in the neural network’s classification capabilities, and vice versa.
The Biologix solution has already passed several tests and undergone continuous refinement by a dedicated research and innovation team. Today, after multiple versions, the accuracy in detecting and classifying snoring reaches 92%. Finally, it is important to note that because this process is 100% automated, the privacy of the patient’s audio is fully protected. Only the patient and their healthcare provider have access to the data collected during the Biologix Sleep Test®.
References
Campos AI, García-Marín LM, Byrne EM, Martin NG, Cuéllar-Partida G, Rentería ME. Insights into the aetiology of snoring from observational and genetic investigations in the UK Biobank. Nat Commun. 2020 Feb 14;11(1):817. doi: 10.1038/s41467-020-14625-1. PMID: 32060260; PMCID: PMC7021827.
Yaremchuk K. Why and When to Treat Snoring. Otolaryngol Clin North Am. 2020 Jun;53(3):351-365. doi: 10.1016/j.otc.2020.02.011. Epub 2020 Apr 24. PMID: 32336469.