Enhanced Feature Extraction and Integration with Advanced Machine Learning Algorithms for Sports-Injury CT-Image Classification
5 affiliations
- 1General Education Department, Anhui Xinhua University, Hefei 230088, China
- 2School of Mathematics-Physics and Finance, Anhui Polytechnic University, Wuhu 241000, China
- 3College of Computer Science and Artificial Intelligence, Wenzhou University, Wenzhou 325035, China
- 4School of Information and Communication Engineering, University of Electronics Science and Technology, Chengdu, China
- 5CenBRAIN Neurotech Center of Excellence, School of Engineering, Westlake University, Hangzhou, China
Abstract
Introduction: X-rays, Computed Tomography (CT) scans, Magnetic Resonance Imaging (MRI), and ultrasound scans are examples of common medical imaging methods. These techniques are highly effective at extracting features, classifying, and diagnosing bone and muscle injuries. Enhancing athletic performance, minimizing injuries, and promoting players' overall health are the main goals of the vital discipline of sports medicine. Sports feature extraction of medical images, and sports medicine specialists help athletes reach their full potential while preserving their long-term health by preventing injuries and providing precise diagnoses.
Method: We proposed a feature extraction approach for sport CT inquiry images, which were first preprocessed and then subjected to wavelet multi-scale analysis. The CT scan image of a sports injury is broken down into different directions based on smooth and rough regions. The Non-Down-Sampled Contour Transformation (NSCT) and Pyramid Bank (NSPFB) model is applied to the CT image of a sports injury. Furthermore, we employed Feedforward Neural Networks (FNNs), Convolutional Neural Networks (CNNs), and recurrent neural networks (RNNs) for medical image diagnosis. The model was tuned by using hyperparameters.
Result: We proposed a feature-extraction model for medical CT scans of sports injuries and provided supporting statistics for medical diagnosis. The CT scan image of the sports injury is subdivided into different sub-directions based on smooth and rough regions. Furthermore, enhanced sub-regain is achieved using an adaptive gamma correction and multi-directional image enhancement via wavelet layering. Our proposed method has demonstrated a remarkable accuracy of 88%, while the deep learning model TabNet-Transform has achieved an imposing 90%. The proposed algorithm improves the quality of the output images and is computationally efficient.
Discussion: Current research indicates that the most vital area of medicine is image analysis, which helps radiologists to make more accurate diagnoses. The feature extraction process involves extracting valuable information from higher-level representations, such as the Region of Interest (ROI) for image analysis, and transforming the image from these representations into lower-level pixel data. The performance analysis of the proposed scheme yielded an efficient extraction time. When extracting feature characteristics for 10 to 30 medical CT images of sports injuries, the average computation time is 2ms to 6 ms.
Conclusion: This research explored the application of machine learning and advanced artificial intelligence models to predict disease diagnosis and to enhance feature extraction from medical CT images of sports injuries. However, several variables impact the original image, including removing image noise and detecting specific regions or objects of interest. Our findings pave the way for further investigations of sports injury and early disease prognosis. We may incorporate data from wearable devices to improve their predictive power and extract the desired features for disease prognosis analysis.
Related
- Optimization-enhanced machine failure classification using critical sensor features and hybrid learning models with advanced optimization techniques
- Integration of Immune checkpoint inhibitors within the definitive radiotherapy-based treatment continuum for locally advanced mucosal head and neck squamous cell carcinoma: a scoping review protocol
- Multi-task deep learning framework for brain tumor classification and segmentation using ensemble feature extraction
- Hybrid feature pyramid attention and capsule network (FpaNet+CapsNet) for enhanced spatial and contextual feature extraction for improved brain tumor segmentation
- Comparison of bone repair of advanced platelet-rich fibrin (A-PRF) and leukocyte- and platelet-rich fibrin (L-PRF) following the extraction of impacted third molars: a randomized clinical trial.
BibTeX
@article{jun2026enhanced,
title = {{Enhanced Feature Extraction and Integration with Advanced Machine Learning Algorithms for Sports-Injury CT-Image Classification}},
author = {Wang Jun and Chu Huiqin and Rashid Abbasi and Md Belal Bin Heyat},
journal = {Current Medical Imaging},
year = {2026},
month = sep,
doi = {10.2174/0115734056413312251111043047},
url = {https://doi.org/10.2174/0115734056413312251111043047},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Enhanced Feature Extraction and Integration with Advanced Machine Learning Algorithms for Sports-Injury CT-Image Classification
AU - Wang Jun
AU - Chu Huiqin
AU - Rashid Abbasi
AU - Md Belal Bin Heyat
PY - 2026
DA - 2026/09/22
JO - Current Medical Imaging
DO - 10.2174/0115734056413312251111043047
UR - https://doi.org/10.2174/0115734056413312251111043047
AB - Introduction: X-rays, Computed Tomography (CT) scans, Magnetic Resonance Imaging (MRI), and ultrasound scans are examples of common medical imaging methods. These techniques are highly effective at extracting features, classifying, and diagnosing bone and muscle injuries. Enhancing athletic performance, minimizing injuries, and promoting players' overall health are the main goals of the vital discipline of sports medicine. Sports feature extraction of medical images, and sports medicine specialists help athletes reach their full potential while preserving their long-term health by preventing injuries and providing precise diagnoses. Method: We proposed a feature extraction approach for sport CT inquiry images, which were first preprocessed and then subjected to wavelet multi-scale analysis. The CT scan image of a sports injury is broken down into different directions based on smooth and rough regions. The Non-Down-Sampled Contour Transformation (NSCT) and Pyramid Bank (NSPFB) model is applied to the CT image of a sports injury. Furthermore, we employed Feedforward Neural Networks (FNNs), Convolutional Neural Networks (CNNs), and recurrent neural networks (RNNs) for medical image diagnosis. The model was tuned by using hyperparameters. Result: We proposed a feature-extraction model for medical CT scans of sports injuries and provided supporting statistics for medical diagnosis. The CT scan image of the sports injury is subdivided into different sub-directions based on smooth and rough regions. Furthermore, enhanced sub-regain is achieved using an adaptive gamma correction and multi-directional image enhancement via wavelet layering. Our proposed method has demonstrated a remarkable accuracy of 88%, while the deep learning model TabNet-Transform has achieved an imposing 90%. The proposed algorithm improves the quality of the output images and is computationally efficient. Discussion: Current research indicates that the most vital area of medicine is image analysis, which helps radiologists to make more accurate diagnoses. The feature extraction process involves extracting valuable information from higher-level representations, such as the Region of Interest (ROI) for image analysis, and transforming the image from these representations into lower-level pixel data. The performance analysis of the proposed scheme yielded an efficient extraction time. When extracting feature characteristics for 10 to 30 medical CT images of sports injuries, the average computation time is 2ms to 6 ms. Conclusion: This research explored the application of machine learning and advanced artificial intelligence models to predict disease diagnosis and to enhance feature extraction from medical CT images of sports injuries. However, several variables impact the original image, including removing image noise and detecting specific regions or objects of interest. Our findings pave the way for further investigations of sports injury and early disease prognosis. We may incorporate data from wearable devices to improve their predictive power and extract the desired features for disease prognosis analysis.
SN - 1573-4056
ER -