Related Experiment Video
Updated: May 26, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.6K
Optimizing Bi-LSTM networks for improved lung cancer detection accuracy
Su Diao1, Yajie Wan2, Danyi Huang3
1Department of Industrial & Systems Engineering, Auburn University, Auburn, Alabama, United States of America.
Plos One
|February 24, 2025
Summary
This study compared hand-crafted and deep learning methods for lung cancer detection. Deep learning, specifically a Bidirectional Long Short-Term Memory (Bi-LSTM) network, achieved superior accuracy for early lung cancer diagnosis.
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Healthcare
- Oncology
Background:
- Lung cancer is a major global health concern, with late diagnosis significantly impacting survival rates.
- Computer-aided diagnosis (CAD) systems utilizing medical image feature extraction show promise but face challenges in identifying optimal features.
- Accurate and early detection of lung cancer is crucial for improving patient outcomes.
Purpose of the Study:
- To compare the diagnostic performance of hand-crafted image features versus deep learning approaches for lung cancer detection.
- To evaluate the effectiveness of Gray Level Co-occurrence Matrix (GLCM) features with Support Vector Machine (SVM) against a Bidirectional Long Short-Term Memory (Bi-LSTM) network.
- To determine the optimal methodology for enhancing lung cancer diagnosis systems.
Main Methods:
- Extraction of traditional hand-crafted features, including Gray Level Co-occurrence Matrix (GLCM) features.
- Application of traditional machine learning algorithms, specifically Support Vector Machine (SVM) with various kernels.
- Optimization and implementation of a deep learning model, Bidirectional Long Short-Term Memory (Bi-LSTM) network, for lung cancer detection.
Main Results:
- Hand-crafted GLCM features combined with SVM achieved high performance, with an accuracy of 99.78% and an Area Under the Curve (AUC) of 0.999.
- The Bidirectional Long Short-Term Memory (Bi-LSTM) deep learning network demonstrated superior performance, reaching an accuracy of 99.89% and an AUC of 1.0000.
- Deep learning models significantly outperformed traditional methods in lung cancer diagnosis accuracy.
Conclusions:
- Deep learning approaches, particularly the Bi-LSTM network, offer enhanced capabilities for lung cancer detection compared to traditional methods.
- Combining hand-crafted features with deep learning shows significant potential for improving the accuracy and effectiveness of lung cancer diagnosis systems.
- The proposed methodologies contribute to advancing early lung cancer detection and improving patient survival rates.
