Classification of transcranial Doppler signals using artificial neural network

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Transcranial Doppler (TCD) is crucial for assessing cerebral blood flow.
• Fast Fourier Transform (FFT) analysis of TCD signals can lead to misinterpretations, especially with turbulent flow.
• Accurate diagnosis of cerebrovascular conditions requires reliable TCD signal analysis.

Purpose of the Study:

• To enhance the accuracy and speed of diagnosing cerebrovascular conditions using TCD signals.
• To investigate the efficacy of artificial neural networks (ANNs) in classifying and analyzing TCD data.
• To compare the performance of different ANN algorithms for TCD signal processing.

Main Methods:

• TCD signals from 110 patients were recorded and digitized.
• Fast Fourier Transform (FFT) was initially applied, revealing limitations in spectral resolution.
• Artificial neural network models, including Backpropagation and Self-Organizing Maps, were trained using Momentum and Delta-Bar-Delta learning algorithms.

Main Results:

• ANNs demonstrated improved classification of TCD blood flow signals compared to traditional FFT methods.
• The study successfully compared the performance of different ANN training and learning algorithms.
• The findings indicate ANNs offer a more robust approach to TCD signal interpretation.

Conclusions:

• Artificial neural networks provide a more accurate and reliable method for analyzing TCD signals.
• ANNs can overcome the spectral resolution limitations of FFT in TCD analysis.
• This approach holds promise for improving the diagnosis of cerebrovascular diseases.