Home ยป Can Parkinson's Disease Predict by Gait?

Can Parkinson's Disease Predict by Gait?

by alex

Can Parkinson's Disease Predict by Gait? Scientists from Osaka University have proposed a neural network that can identify Parkinson's disease based on physical activity. Moreover, the network can detect the problem both in humans and in mice, beetles and even roundworms.

Bioinformatics from Japan have developed a neural network capable of detecting motor traits of Parkinson's disease. Moreover, movement disorders, potentially indicating the development of such a disease, are determined by the network not only in humans, but also in mice, beetles and worms. A study of the operation of a neural network was published in the journal Nature Communication.

As a basis for their development, scientists took the fact that with various diseases of a neurodegenerative nature – be it Parkinson's disease, Alzheimer's disease, schizophrenia and others – there is a problem with movement. Failure develops due to a violation of dopamine production – the neurotransmitter is not only part of the self-reward system and the hormone of joy, but is also responsible for the regulation of the body's motor activity. Dopamine is closely associated with the neurons of the substantia nigra. If they die, the body loses coordination, cannot move normally, against the background of which tremor develops, as well as muscle overstrain and akinesia (the impossibility of voluntary movements or their change in strength, volume, speed due to paralysis, pain or immobility of the joints ).

Model organisms are used to study various internal processes – after all, it is safe to control various parameters, such as the amount of dopamine in a person, it is difficult. Therefore, evolutionarily close models, for example, the same mice, are carefully studied. True, scientists are faced with a problem here – a person and models move in different ways. And the results cannot be interpreted simply by comparison.

Japanese scientists have created a model that defines the general patterns of movement for different models – humans, mice, beetles and roundworms. It is presented in the form of a domain-adversarial neural network. It receives data on the parameters of movement, in particular, speed and trajectory, and then displays data about the domain – healthy or with the presence of Parkinson's disease. Moreover, the conclusion is given by the domain – a person, a mouse or another organism.

To determine the general patterns of the disease, it is necessary to extract those variants that the neural network identifies by class, but cannot recognize by type. The quality of interpretation of the received data in the neural network is provided by a special attention module. All data is provided in the form of a graph. The neural network was trained on different pairs to find common features. For example, although the worm and the mouse move differently, the same trait is defined: the inability to maintain a high speed of movement for a long time.

In a pair of man and a worm in Parkinson's disease, a characteristic feature is determined – an unstable speed during acceleration. In a pair, the worm and the beetle both had the same pattern as the inability to make smooth turns. All these options, scientists say, can be used in the future to study the amount of dopamine, as well as use the neural network to analyze other motor disorders.

The following sources were used to prepare the material:

Nature Communication

You may also like

Leave a Comment