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

Diagnostics, 2023 · DOI: 10.3390/diagnostics13081417 · Published: April 14, 2023

Neuroimaging Medical Imaging Bioinformatics

Simple Explanation

The spinal cord is segmented into five regions, and these segments are used to train CNN classifiers. Each classifier is responsible for detecting a particular type of tumor, which assists in improving model scalability and performance. The model segments all five spinal cord regions and stores them as separate datasets manually tagged with cancer status.

Study Duration

Not specified

Participants

5000 images from Mendeley datasets

Evidence Level

Not specified

Key Findings

1
The proposed model achieved a 14.5% better segmentation efficiency compared to state-of-the-art models.
2
The model achieved 98.9% tumor classification accuracy.
3
The model has a 15.6% higher speed performance when averaged over the entire dataset.

Research Summary

This paper proposes a novel augmented model for spinal cord segmentation and tumor classification using deep nets to overcome this limitation. Multiple Mask Regional Convolutional Neural Networks (MRCNNs) were trained on various datasets for region segmentation. Due to use of specialized CNN models for different spinal cord segments, the proposed model was able to achieve a 14.5% better segmentation efficiency, 98.9% tumor classification accuracy, and a 15.6% higher speed performance.

Practical Implications

Improved Diagnostics

The model can improve the accuracy and efficiency of spinal cord tumor diagnosis.

Clinical Deployment

The model's performance makes it suitable for various clinical deployments.

Scalability

The model is highly scalable for a wide variety of spinal cord tumor classification scenarios.

Study Limitations

1
The models were developed for particular portions of the spine
2
Verify the performance of this model on other datasets
3
replace CNN models with recurrent NN (RNN) models to further improve classiﬁcation the capabilities for larger datasets

Your Feedback

Was this summary helpful?

Back to Neuroimaging