A Novel Hybridized Feature Extraction Approach for Lung Nodule Classification Based on Transfer Learning Technique.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Deep learning models, particularly Convolutional Neural Networks (CNNs), show promise in medical diagnosis.
• Traditional handcrafted features may not fully capture complex nodule characteristics.
• Large annotated datasets are often required for training deep learning models.

Purpose of the Study:

• To develop a hybridized model for accurate lung nodule classification (benign vs. malignant).
• To integrate deep features from Residual Neural Network (ResNet) with handcrafted Histogram of Oriented Gradients (HOG) features.
• To overcome limitations of solely relying on handcrafted or deep features.

Main Methods:

• A hybrid model was created by combining ResNet-derived deep features and HOG features.
• A Radial Basis Function Support Vector Machine (RBF-SVM) was employed for classification.
• The model was evaluated using the LIDC-IDRI dataset.

Main Results:

• The hybridized model achieved high performance metrics: 97.53% accuracy, 98.62% sensitivity, and 96.88% specificity.
• Precision reached 95.04%, with an F1 score of 0.9679.
• False-positive rate was 3.117% and false-negative rate was 1.38%.

Conclusions:

• The proposed hybridized feature-based classification technique outperforms deep feature-based methods for lung nodule classification.
• This hybrid approach offers a robust solution for improving lung cancer diagnosis.