Researchers from the Hospital Universitario Severo Ochoa and the Universidad Carlos III de Madrid (UC3M) have collaborated on a study that applies artificial intelligence to the study of brain electrical activity during sleep, with the goal of developing tools that contribute to the early detection of Alzheimer’s disease. According to a study published in GeroScience, machine learning algorithms can identify incipient neuronal alterations and categorize patients into three distinct biological subgroups based on nocturnal brain waves.
How AI Uses Sleep Patterns to Identify Alzheimer’s
The research focuses on the sleep period during which the brain performs cellular cleaning processes of metabolic waste, including beta-amyloid protein.
The team used machine learning to analyze polysomnographies—records of brain activity captured via electrodes on the scalp. This non-invasive approach allows the AI to detect specific changes in electrical signals that may detect the accumulation of proteins in the brain. Anna Michela Gaeta, a pulmonology specialist at Hospital Universitario Severo Ochoa, stated that the goal is to create a diagnostic tool that is economical, non-invasive, and can be extended to the majority of the population.
Biomarker Correlation and Patient Subgrouping
The study analyzed a dataset of 42 patients with mild-to-moderate Alzheimer’s and 58 cognitively healthy controls. This data was sourced from a previous repository created by Gerard Piñol Ripoll of Hospital Universitario Santa María de Lérida-IRBLleida and Ferran Barbé of Hospital Universitario Arnau de Vilanova-IRBLleida, with funding from the Instituto de Salud Carlos III.
By crossing AI-analyzed sleep data with cerebrospinal fluid biomarkers, the researchers identified three distinct “subclusters” of Alzheimer’s patients. The AI tracked variations in the following key markers:
- Beta-amyloid (A42): Biomarker of the cerebrospinal fluid.
- Phosphorylated tau (p-tau181): Biomarker of the cerebrospinal fluid.
- Total tau (t-tau): Biomarker of the cerebrospinal fluid.
- Neurofilament light chain (NfL): Biomarker of the cerebrospinal fluid.
Comparing Current Diagnostic Methods
Current Alzheimer’s diagnostics often require complementary information through advanced techniques or invasive procedures that are typically performed in relatively late stages. The following table contrasts the AI sleep analysis approach with traditional methods described in the study:
| Method | Invasiveness | Typical Timing | Accessibility |
|---|---|---|---|
| AI Sleep Analysis | Non-invasive | Future potential for early detection | Potential for home use |
| Lumbar Puncture | Invasive | Relatively late stages | Clinical Setting |
| PET Scan | Advanced technique | Relatively late stages | Clinical Setting |
| p-tau217 Blood Test | Plasma determination | Early phases | Limited availability in Spanish hospitals |
Clinical Implications and Future Application
Because Alzheimer’s symptoms usually manifest between 10 and 20 years after the start of pathological processes in the brain, early detection is critical. Dr. Gaeta noted that current pharmaceuticals are only effective if administered in the earliest stages of the disease.
The researchers envision a future where AI-driven sleep monitoring complements blood tests, such as those for the p-tau217 protein. This combined approach could create a low-cost screening pathway to detect preclinical Alzheimer’s and treat comorbid sleep disorders, which may help slow the progression of cognitive decline. The study emphasizes that future science in this field will require the integration of neurology, pulmonology, and engineering.
Keep reading