ELectronic Archimedes spiral Neural Network (ELANN)

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Imaging

Background:

• The Archimedes spiral is a diagnostic tool for essential tremor.
• Subjectivity in spiral rating by physicians can affect diagnostic consistency.
• Need for an objective, standardized method for tremor evaluation exists.

Purpose of the Study:

• To develop an objective standard model for essential tremor evaluation using convolutional neural networks.
• To assess the model's performance against experienced neurologists in rating Archimedes spirals.

Main Methods:

• A convolutional neural network was trained on 1653 Archimedes spiral images rated by a senior neurologist.
• The model's performance was compared with ratings from two senior neurologists on a test dataset.
• Inter-rater reliability was quantified using Cohen's Weighted Kappa.

Main Results:

• The AI model achieved high agreement with the reference standard (Kappa 0.89-0.96).
• The model's performance was comparable to two senior movement disorder neurologists (Kappa 0.86-0.98).
• Overlapping confidence intervals indicate similar proficiency between the model and expert raters.

Conclusions:

• The developed convolutional neural network model objectively evaluates Archimedes spirals with proficiency similar to experienced neurologists.
• This AI tool offers a standardized and reliable method for essential tremor assessment.
• The model has the potential to improve diagnostic consistency and accuracy in essential tremor monitoring.