Trends in brain MRI and CP association using deep learning.

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Cerebral palsy (CP) is a neurological disorder impacting motor function and quality of life.
• Early diagnosis of CP is challenging due to infant uncooperativeness and limitations in current imaging analysis.
• Timely identification of CP through brain MRI is vital for effective intervention and management.

Purpose of the Study:

• To introduce novel deep learning models for enhanced cerebral palsy detection in brain MRIs.
• To investigate the efficacy of single-sequence and multiple MRI scans for CP identification.
• To develop models capable of identifying vulnerable brain regions and sensitive MRI slices associated with CP.

Main Methods:

• Development and training of two deep learning models: SSeq-DL (single-sequence) and SMS-DL (multiple-sequence MRIs).
• Incorporation of specialized attention mechanisms, parallel computing, and layer-wise fusion for enhanced feature learning.
• Experimentation with single and coupled MRI scans to assess model performance and identify critical imaging features.

Main Results:

• Both SSeq-DL and SMS-DL models demonstrated significant capability in identifying cerebral palsy.
• The models successfully highlighted lesion-vulnerable regions and sensitive MRI slices associated with CP.
• Analysis revealed trends in affected slices across different age ranges, aiding in early detection.

Conclusions:

• Deep learning models offer a promising approach for early and accurate cerebral palsy detection using brain MRIs.
• The study validates the utility of both single and multiple MRI sequences for CP identification.
• Findings support the use of these models to assist radiologists in identifying early signs of CP and guiding rehabilitation efforts.