Related Experiment Video
Updated: Aug 6, 2026
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
A Hybrid CNN-Transformer Model with Crossover Boosted Cheetah Optimization for Prenatal Spina Bifida Identification
Asha Rajasekaran1, S S Subashka Ramesh2
1Department of CSE, Sathyabama Institute of Science and Technology, Chennai, India.
Summary
This study introduces a novel deep learning model for accurate spina bifida detection in prenatal ultrasounds. The hybrid model achieves high accuracy, offering a reliable tool for early identification of this neural tube defect.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Obstetrics
Background:
- Spina bifida is a neural tube defect impacting spinal cord development.
- Existing deep learning models struggle with fine details in ultrasound images, limiting spina bifida detection accuracy.
- Accurate, automated detection of spina bifida from second-trimester ultrasounds is crucial for prenatal care.
Purpose of the Study:
- To develop a trustworthy, automated deep learning framework for precise spina bifida identification.
- To enhance the extraction of local and global features from fetal ultrasound images for improved diagnostic accuracy.
- To create a robust model for real-time prenatal screening.
Main Methods:
- A novel hybrid convolutional neural network-transformer model was proposed.
- Multi-resolution CNN blocks and positional encoding enhanced spatial feature representation.
- A self-attention mechanism highlighted clinically relevant regions, and hyperparameters were optimized using a crossover boosted cheetah optimization algorithm.
Main Results:
- The model achieved a 99.02% accuracy and a 98.21% F1-score.
- Demonstrated a low inference time of 0.88 seconds, suitable for real-time applications.
- The proposed model showed robust diagnostic performance for spina bifida identification.
Conclusions:
- The developed model offers an efficient and clinically viable solution for automated prenatal spina bifida detection.
- This deep learning approach can significantly aid in early and accurate diagnosis of spina bifida.
- The framework shows promise for integration into routine prenatal ultrasound screening.