Specialists from the Immanuel Kant Baltic Federal University have presented a revolutionary development in the field of medical diagnostics. The multimodal neural network they created is capable of detecting brain function disorders with high accuracy by analyzing only an audio recording of a patient's speech. This is not just another recognition algorithm—it is a full-fledged tool for primary screening of cognitive disorders.

How does the technology work?

Cognitive impairments, whether early stages of dementia or other pathologies of the nervous system, often masquerade as ordinary fatigue or age-related changes in their early phases. Traditional diagnostics require time and significant financial costs. However, a person's voice changes long before clinical symptoms become obvious. It is this pattern that the new neural network utilizes.

The system analyzes not only the semantic content of the patient's response (for example, describing a picture) but also 88 acoustic parameters: intonation, pauses, speech tempo, voice tremor, and volume fluctuations. Specialized medical tests were not used to train the model—the main emphasis was placed precisely on acoustic features. The neural network compares the obtained data with a database of speech samples from both healthy individuals and patients with confirmed diagnoses, then produces a result on a scale from 0 to 100 percent, indicating the probability of pathology.

Prospects and limitations

It is important to emphasize: the development does not make final diagnoses. It is a powerful tool for primary screening that will become an indispensable assistant for doctors, allowing them to make clinical decisions faster. The team is already launching joint projects with leading hospitals to collect a large-scale database of voice recordings, which is necessary to transform the prototype into a full-fledged medical product. The research results were presented at the XV International Multiconference on Systems Biology in Novosibirsk.

Expert opinion: The integration of AI into medicine is advancing by leaps and bounds, and voice diagnostics is one of the most promising areas. The ability to conduct fast and inexpensive screening of cognitive disorders at an early stage could radically change the approach to treating dementia and other neurodegenerative diseases. The next step is ensuring accuracy and data confidentiality, which is critically important for implementation in clinical practice.