Using ventricular modeling to robustly probe significant deep gray matter pathologies: Application to cerebral palsy

Area of Science:

• Neuroimaging
• Quantitative MRI analysis
• Cerebral Palsy research

Background:

• Assessing deep gray matter injury in Cerebral Palsy (CP) is challenging due to significant anatomical variations.
• Current methods often rely on qualitative Magnetic Resonance Image (MRI) assessments or standard segmentation, which can be unreliable with large injuries.
• Quantitative measures are needed to support clinical assessments and track functional outcomes.

Purpose of the Study:

• To develop a robust surrogate marker for deep gray matter injury in children with CP.
• To quantify injury extent based on local ventricular enlargement impacting surrounding anatomy.
• To correlate this novel measure with clinical functional outcomes.

Main Methods:

• Constructed a statistical shape model of lateral ventricles from 44 healthy subjects.
• Quantified local ventricular enlargement as a surrogate for deep gray matter injury.
• Trained a regression model using data from 95 CP patients to predict clinical function.
• Validated the method against expert clinical assessment and standard segmentation approaches.

Main Results:

• The automated method achieved an area under the curve of 0.91 in identifying ventricular enlargement against expert assessment.
• The surrogate marker showed strong significant correlations with motor function (r²=0.62), executive function (r²=0.55), and communication (r²=0.50).
• This approach demonstrated superior performance compared to standard anatomical segmentation methods, especially with large anatomical variations.

Conclusions:

• The proposed automated method provides a robust and reliable marker for deep gray matter injury in CP.
• Its independence from precise anatomical segmentation makes it suitable for cases with significant brain abnormalities.
• This technique shows significant potential for routine MRI assessment in children with CP, aiding in diagnosis and monitoring.