Early detection of dementia disease that can save lives.
Selected examples of recognition:
Dementia diseases are not just a medical challenge of the 21st century - they are also a global issue of economic, social, and research significance. They affect the quality of life of millions of people, leading to a loss of independence and requiring significant financial outlays for care.
Currently, over 50 million people suffer from dementia. Experts predict an increase to 152 million by 2050.
Source: World Alzheimer Report 2018
While Alzheimer's disease is dominant, it accounts for 60-70% of cases. There are also other types of dementia with equally serious effects.
Source: WHO
Dementia diseases affect patients, their families, caregivers, and the entire society. It's time for global action to support and educate them.
Source: World Alzheimer Report 2018
Global costs
$818 billion
in 2019.
$1.1 trillion
by 2030.
Source: WHO
Costs in Europe
€170 billion
in 2020.
€250 billion
by 2030.
Source: ECDC
Sound Objects is Vivid Mind's unique, patented approach to audio decomposition, separation and analysis using sound vectoring methods and algorithms. This advanced signal processing methodology is based on a sinusoidal model that offers a radically new approach to computational analysis of the auditory scene (i.e., machine listening). Read the scientific publication "Screening method for early dementia using sound objects as voice biomarkers."
In current advanced spectral editors, the audio track is represented on the spectrogram as a heat map of points (pixels/raster graphics), which is challenging to semantically group and manipulate.
In the Vivid Mind solution, the energy points visible on the spectrogram are automatically grouped and condensed into so-called sound objects. These are graphical three-dimensional vector representations of all frequency components, where each condensed energy point is connected with a phase sequence (a unique and essential feature of Vivid Mind technology).
Phase is one of the most crucial elements of speech analysis. It allows for the analysis not only of individual energy points but also their alignments, enabling focus on specific harmonic sounds or phonemes.
The technology using phase is one of the most sensitive predictors of cognitive disorders. It allows for the detection of irregularities in speech control, such as disturbances or phase jumps.
In speech analysis, phase is defined as the relationship between two points in time. It can be used to analyze changes in sound, such as frequency shifts or amplitude.
Technology that utilizes phase in speech analysis can be used for various purposes, such as:
The Sound Objects technology is revolutionizing the approach to early diagnosis of dementia, offering the possibility of identifying initial symptoms based on patient voice analysis. This innovative tool can redefine the future of medicine, allowing for earlier and more precise diagnosis and monitoring of disease progression.
Recording the Patient's Voice: A 10-second voice sample is taken from the patient.
Voice Transformation
Data Transmission to Classifier: The extracted data is then transmitted to a classifier.
Based on the data analysis, it’s recommended either to visit a specialist or repet regular check-ups at a general practitioner's clinic.
n a "blind" study conducted in the summer of 2022 to test the early performance of the VM technology, carried out by Professor Jan Konopacki from the Neurological Clinic of the Medical University in Łódź, Poland, we received 33 recordings suitable for analysis. Out of these, 5 recordings were not analyzed due to their poor technical quality. Our detection system recorded the following parameters.
The percentage of sick individuals correctly identified by the system.
100%
The percentage of healthy individuals correctly identified as not having the disease by the system.
81%
The percentage of individuals correctly identified by the system, whether they have the disease or not.
82%
The percentage of individuals identified by the system as sick who are actually sick.
50%
The system's resistance to errors and noise in the data.
In progress
The smallest change in voice that the system can detect.
In progress
The smallest change in voice that the system can quantify.
In progress
Our detection system is still in the development phase, and its parameters may change in the future.
Advanced audio objects provide a compact, yet precise representation of the human voice spectrogram.
Working in conjunction with artificial intelligence technologies, they enable us to detect subtle nuances in the voice that may indicate the onset of neurodegenerative diseases.
The central element of our approach is the system's ability to continuously improve. Our algorithm enhances its precision with each analysis, which means it can detect cognitive disturbances even at a very early stage.
In a world where diagnostic errors can have serious consequences, trust in diagnostic tools is crucial. Hence, the transparency and understandability of our AI models are our priority.
We want every doctor to be sure of how our algorithm works and on what basis it makes decisions.Our plans to enhance our innovative edge involve continuing actions in the field of intellectual property protection, including the registration of new patents. Additionally, we aim to increase the precision and accuracy of our algorithms. This will primarily be achieved by rapidly expanding our data set used for training machine learning (ML) algorithms.
The precision of the sound spectrum in Vivid Mind technology surpasses competitive solutions for diagnosing dementia based on voice analysis. While other methods focus mainly on the semantic layer and various aspects of speech, Vivid Mind concentrates on the physiological aspects of sound articulation, such as harmonic frequency, noise, vibration, and tremor.
Thanks to the analysis of over 100 million data points from a short recording (less than 10 seconds), Vivid Mind is able to detect subtle patterns of speech changes characteristic for patients with dementia.
As part of the collaboration with the Medical University of Wrocław, our tool's model was tested on a sample of 250 voice recordings of individuals with neurodegenerative diseases. Research showed that the model can detect the presence of the disease with high accuracy (95%).
The research results have been published in a pre-validation report, which provides detailed information about the research methodology, findings, and their interpretation.
Vivid Mind is conducting an experiment in partnership with the Pomeranian Medical University in Szczecin. The goal is to develop an effective, simple, and inexpensive method for preliminary assessment of cognitive functions. The experiment involves 750 people over the age of 50 and 250 healthy individuals aged 18–30.
In progress
Vivid Mind is fully compliant with the General Data Protection Regulation (GDPR). They guarantee that your personal data is safe and treated with due diligence.
Compliance with GDPR
EN
PL
English