Enhancing lung cancer detection through hybrid features and machine learning hyperparameters optimization techniques.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Early lung cancer detection is critical for improving patient survival rates.
• Accurate feature extraction is a key challenge in machine learning for medical diagnosis.
• Combining relevant features can significantly boost the performance of diagnostic models.

Purpose of the Study:

• To develop and evaluate a hybrid feature extraction method for lung cancer detection.
• To assess the effectiveness of integrating Gray-level co-occurrence matrix (GLCM) with autoencoder features.
• To improve the accuracy of supervised machine learning models in identifying lung cancer.

Main Methods:

• A hybrid feature extraction approach combining GLCM with Haralick features and autoencoder features was developed.
• These integrated features were utilized as input for supervised machine learning algorithms.
• Support Vector Machine (SVM) models, including Radial Base Function (RBF), Gaussian, and polynomial kernels, were employed for classification.

Main Results:

• The hybrid approach integrating GLCM, Haralick, and autoencoder features achieved high accuracy with SVM polynomial (99.89%).
• SVM Gaussian and SVM RBF demonstrated perfect performance measures using the combined feature set.
• SVM Gaussian achieved 99.56% accuracy and SVM RBF achieved 99.35% accuracy using GLCM with Haralick features.

Conclusions:

• The proposed hybrid feature extraction method shows significant potential for enhancing lung cancer detection accuracy.
• This approach can contribute to the development of improved systems for lung cancer diagnosis and treatment planning.
• Machine learning, particularly with advanced feature integration, offers a promising avenue for advancing oncological diagnostics.