Russian scientists from Immanuel Kant Baltic Federal University have presented an innovative development—a multimodal neural network capable of detecting cognitive impairments from audio recordings of speech. This technology is designed for the primary diagnosis of patients suspected of having dementia and other nervous system pathologies at the earliest stages.
Many cognitive disorders, such as memory decline or reduced mental performance, can easily be mistaken for ordinary fatigue or age-related changes in their early stages. However, it is precisely at this point that changes in the voice become one of the first and most informative markers. Traditional diagnostic methods require lengthy and expensive examinations, whereas voice analysis can be conducted quickly and with minimal costs.
How the algorithm works
The developed artificial intelligence model processes an audio file, simultaneously analyzing the semantic content of the speech and 88 acoustic parameters. Unlike standard testing, where a doctor evaluates a patient's response to describing a picture, the neural network captures the finest nuances: voice tremor, pause duration, volume fluctuations, and speech tempo.
The system compares the obtained data with an extensive database of samples—both from healthy individuals and patients with confirmed diagnoses. The final result is presented as a probability of pathology on a scale from 0 to 100%. Notably, no specialized medical tests were used in training the neural network—the focus was solely on acoustic features.
Comparison with traditional examination:
- Basis of assessment: Traditional method—content of the patient's response. Neural network—text meaning + 88 acoustic parameters.
- Speed of operation: Traditional method—lengthy doctor's appointment. Neural network—automatic instant processing.
- Financial costs: Traditional method—high. Neural network—minimal.
Limitations and prospects
It is important to emphasize: the development is not intended to make an accurate diagnosis. It is a tool for primary screening that will serve as a convenient aid for doctors in clinical decision-making. The team plans to turn the technology into a full-fledged medical product. This will require collecting a large-scale database of voice recordings, and 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.
My comment: This is a vivid example of how AI is penetrating areas requiring high precision and empathy. For the crypto industry, where we are accustomed to data analysis and automation, such developments serve as a reminder that technology can solve not only financial but also socially significant problems. The implementation of such tools in healthcare could drastically reduce the burden on the system and improve the quality of life for millions of people.