Analysis of speech patterns is one of the most promising non-invasive methods for early detection of neurodegenerative diseases. A group of researchers from Immanuel Kant Baltic Federal University has presented a multimodal neural network capable of identifying cognitive impairments with high accuracy from audio recordings of speech.

The development is aimed at primary screening of patients. Cognitive disorders such as dementia often masquerade as ordinary fatigue or age-related changes in their early stages. Traditional diagnosis requires lengthy and costly procedures. Meanwhile, acoustic parameters of the voice—intonation, tempo, pauses, tremor—change long before obvious clinical symptoms appear.

How the algorithm works

The system is based on a multimodal artificial intelligence model that simultaneously analyzes the semantic content of speech and 88 acoustic parameters. The neural network processes an audio file, comparing it with a database containing speech samples from both healthy individuals and patients with confirmed diagnoses.

The algorithm identifies key markers of impairments and calculates the probability of pathology on a scale from 0 to 100%. Notably, specialized medical tests were not used during training—the main focus was placed specifically on acoustic features.

Comparison of traditional examination and neural network analysis:

  • Basis of assessment: Traditionally—patient's response content; Neural network—text meaning + 88 acoustic parameters.
  • Speed of operation: Traditionally—lengthy doctor's appointment; Neural network—automatic instant processing.
  • Financial costs: Traditionally—high; Neural network—minimal.

Limitations and plans

It is important to emphasize: the system does not make a final diagnosis. It is designed as a primary screening tool that helps the doctor make a clinical decision faster. The developers plan to turn the technology into a full-fledged medical product. This will require collecting a large-scale database of voice recordings—the university is already launching joint projects with leading hospitals.

The research results were presented at the XV International Multiconference on Systems Biology in Novosibirsk.

Expert opinion: This is exactly the case where AI does not replace the doctor but provides him with a qualitatively new tool. The ability to "hear" the disease at a stage when it has not yet manifested clinically fundamentally changes the approach to the prevention of neurodegenerative diseases. It is especially valuable that the technology does not require complex equipment—a smartphone and a microphone are sufficient.