NAMSTCD: A Novel Augmented Model for Spinal Cord Segmentation and Tumor Classification Using Deep Nets.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Spinal cord segmentation is crucial for diagnosing and treating spinal conditions.
• Existing models often lack scalability, focusing on specific spinal regions.
• This limits their effectiveness in comprehensive spinal cord analysis.

Purpose of the Study:

• To develop a novel augmented deep learning model for accurate spinal cord segmentation and tumor classification.
• To overcome the limitations of existing models by addressing the entire spinal cord.
• To improve the scalability and clinical applicability of spinal cord analysis tools.

Main Methods:

• A novel augmented model was developed using deep neural networks.
• The model segments all five spinal cord regions, creating separate, expert-tagged datasets.
• Multiple Mask Regional Convolutional Neural Networks (MRCNNs) and specialized CNNs (VGGNet 19, YoLo V2, ResNet 101, GoogLeNet) were employed for segmentation and classification.

Main Results:

• The proposed model achieved a 14.5% improvement in segmentation efficiency.
• It demonstrated 98.9% accuracy in tumor classification.
• A 15.6% increase in speed performance was observed compared to state-of-the-art models.

Conclusions:

• The augmented model offers superior performance in spinal cord segmentation and tumor classification.
• Its high accuracy, efficiency, and speed make it suitable for clinical deployment.
• The model's scalability across different tumor types and spinal regions enhances its utility in diverse clinical scenarios.