ShPCFHNet: shepherd parallel convolutional forward harmonic net for spinal cord injury detection using CT images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neurosurgery

Background:

• Computed Tomography (CT) is the primary imaging modality for diagnosing spinal cord injuries (SCI).
• Accurate prediction of functional outcomes in SCI patients relies heavily on early and precise injury diagnosis.
• Current diagnostic methods face challenges in accurately identifying initial clinical injuries.

Purpose of the Study:

• To develop an efficient and accurate model for detecting spinal cord injuries (SCI) from CT images.
• To enhance the prediction of functional outcomes for patients with spinal cord injuries.
• To introduce the Shepard Parallel Convolutional Forward Harmonic Net (ShPCFHNet) for improved SCI detection.

Main Methods:

• CT image enhancement using logarithmic transformations.
• Spinal cord segmentation via a Dual-branch UNet with Sensitivity-Specificity Loss (SSL).
• Disc localization using active contour models, followed by feature extraction and SCI detection with ShPCFHNet (combining ShCNN, PCNN, and Harmonic analysis).

Main Results:

• The ShPCFHNet model achieved high performance metrics.
• Accuracy: 91.397%
• True Positive Rate (TPR): 92.684%
• True Negative Rate (TNR): 90.366%

Conclusions:

• The proposed ShPCFHNet model demonstrates significant potential for accurate SCI detection in CT imaging.
• This AI-driven approach can aid clinicians and radiologists in improving diagnostic accuracy and functional prediction for SCI patients.
• The integration of advanced deep learning and harmonic analysis offers a promising direction for neuroimaging research.