A deep explainable artificial intelligent framework for neurological disorders discrimination.

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning

Background:

• Pathological hand tremor (PHT) is a key symptom in Parkinson's disease (PD) and essential tremor (ET).
• Distinguishing between PD and ET is challenging due to overlapping symptoms, requiring specialized diagnostic methods.
• Accurate diagnosis of PHT is crucial for effective treatment and management.

Purpose of the Study:

• To develop and validate a data-driven model for differentiating between Parkinson's disease and essential tremor.
• To process hand kinematics data for accurate patient classification.
• To provide clinically viable insights into the model's decision-making process.

Main Methods:

• A novel data-driven model, referred to as [Formula: see text], was developed.
• The model was trained on 90 hours of hand motion signals from 250 tremor assessments across 81 patients.
• Machine learning interpretability and explainability measures were employed.

Main Results:

• The [Formula: see text] model achieved exceptional differential diagnosis accuracy, outperforming existing state-of-the-art methods.
• The model demonstrated high accuracy in classifying patients into PD or ET groups.
• Statistically significant and clinically relevant insights were derived from the model's discriminative capabilities.

Conclusions:

• The proposed [Formula: see text] model offers a powerful tool for the differential diagnosis of Parkinson's disease and essential tremor.
• The data-driven approach provides a reliable and accurate method for analyzing pathological hand tremor.
• Explainable AI insights enhance clinical understanding and trust in the diagnostic model.