Imaging Tremor Quantification for Neurological Disease Diagnosis

Area of Science:

• Neurology
• Biomedical Engineering
• Computer Science

Background:

• Tremors are neurological disorders causing involuntary movements, necessitating accurate diagnosis for effective treatment.
• Current diagnostic methods may lack objectivity, highlighting the need for advanced analytical techniques.
• Essential tremor and cerebellar disorders are common types requiring specific diagnostic approaches.

Purpose of the Study:

• To introduce a simple, objective method for tremor assessment and measurement using image analysis and deep learning.
• To develop a hybrid approach for quantifying tremor parameters for disease classification.
• To differentiate between healthy individuals, essential tremor patients, and cerebellar disorder patients.

Main Methods:

• Utilized image processing techniques to analyze test results from a finger-nose-finger test.
• Employed machine learning, specifically support vector machines, for unsupervised clustering of data.
• Grouped data to identify distinct classes of typical tremor responses.

Main Results:

• The proposed method successfully analyzed tremor data using image processing and deep learning.
• Unsupervised clustering using support vector machines resulted in a three-cluster distribution.
• These clusters effectively differentiated between healthy subjects, essential tremor patients, and cerebellar disorder patients.

Conclusions:

• The developed hybrid approach offers an objective and accurate method for tremor assessment.
• The technique shows promise in classifying tremors, aiding in the identification of essential tremor and cerebellar disorders.
• This method provides a foundation for improved diagnostic tools in neurology.