Improved parkinsonism diagnosis using a partial least squares based approach.

Area of Science:

• Neurology
• Medical Imaging
• Machine Learning

Background:

• Diagnosing Parkinsonian syndrome (PS) is challenging due to overlapping symptoms across various conditions.
• Accurate and early diagnosis is crucial for effective patient management.
• (123)I-ioflupane SPECT imaging offers potential for improved in vivo assessment of neurodegenerative disorders.

Purpose of the Study:

• To develop and validate a novel automated method for classifying (123)I-ioflupane brain images.
• To differentiate between controls and patients with Parkinsonian syndrome (PS).
• To enhance the accuracy of PS diagnosis using advanced imaging analysis.

Main Methods:

• A novel classification methodology was developed analyzing each brain hemisphere separately.
• The approach integrates Partial Least Squares (PLS) for feature extraction and Support Vector Machines (SVM) for classification.
• The method was evaluated on a database of 189 (123)I-ioflupane SPECT images.

Main Results:

• The proposed PLS-based method achieved high diagnostic performance.
• Accuracy reached 94.7%, with sensitivity of 93.7% and specificity of 95.7%.
• This method outperformed previous approaches utilizing singular value decomposition.

Conclusions:

• Applying advanced signal analysis techniques to individual brain hemispheres significantly improves assisted diagnosis of PS.
• The developed method offers a robust tool for differentiating PS from healthy controls.
• This technique holds promise for earlier and more accurate detection of Parkinsonian syndromes.