Currently, Neurovisor recognizes signs of tuberculosis and pneumonia, identifying them as a result of x-ray analysis.
Powerful Computer Network
- The system is based on a trainable neural network.
- Its efficiency and speed of learning are provided with a computing power up to 7 teraflops, which match to the 34th place in the world ranking of Top-500 supercomputers.
Neural Network Training
The first version of the system was trained on 10,000 x-ray images.
At the core of the first version of Neurovisor is:
- an artificial intelligence system containing a neural network with ability to segment;
- visual user interface;
- blockchain technology to provide the necessary computing power.
The system capacity is up to 7 teraflops, which is equivalent to 3000 modern office computers.
Data for the Training
For the training of the neural network based on the signs of tuberculosis and pneumonia, open datasets were used:
- National Institutes of Health Chest X-Ray Dataset (USA)
- China Set - The Shenzhen set - Chest X-ray Database (China)
Neurovisor will draw the attention to the foci specific to pneumonia and tuberculosis diseases and will help to make a decision in a non-trivial situation.
After uploading the X-ray image electronically to the Neurovisor, the user receives the recognized information in response. The picture highlights the area where the probable signs of the disease are revealed.
Signs of disease detected.
No signs of disease.
Neurovisor Work Organization
- Time: speeds up the processing of patients' medical data.
- Examination: helps professionals in determining the foci of tuberculosis and pneumonia.
- Reduction of human errors: the system complements the work of a medical specialist, helping to draw attention to potential signs of the disease.
- Assessment: helps to check the effectiveness of radiologists, pulmonologists.
- Accessibility: in the absence of highly qualified specialists in remote areas and regions of a country, the system will help decipher a result and provide a preliminary analysis to any doctor.
- Training: the use of a neural network allows for the further training of artificial intelligence, constantly increasing the accuracy of diagnoses.
- Scalability: the system can be used by both a private specialist and a network of clinics.
- Informational integration: the system is available in boxed version, which can be installed on computer or server of a user. The system does not depend on the connection quality and does not create problems with security of internal networks.
- Updatability: with continuous self-improvement, the system performs periodic updates to improve the accuracy of the result (updates can be disabled as agreed with the user).
- Power: Blockchain technology guarantees the power to train and process vast amounts of medical data.
Privacy by Design
The Neurovisor system is based on the principle of privacy by design. Privacy is a standard setting of the system. Neurovisor is trained and uses impersonal datasets, and also ensures the complete safety of personal data where they are involved.
The underlying algorithms of Neurovisor are aimed at meeting the most stringent requirements for the use of personal data embedded in the GDPR.
Neurovisor is accessible artificial intelligence on a table of any medical professional.
The system complements and optimizes the work of medical professionals. It does not give a 100% result and does not make a diagnosis. However, it gives the necessary clues and a preliminary result, which can be used by a doctor for subsequent diagnosis.