ExtRanFS: An Automated Lung Cancer Malignancy Detection System Using Extremely Randomized Feature Selector

Area of Science:

• Medical Imaging and Artificial Intelligence
• Oncology and Computer-Aided Diagnosis

Background:

• Lung cancer, characterized by uncontrolled cell growth, poses a significant health risk if not detected early.
• Accurate and timely diagnosis of lung tumors as benign, malignant, or normal is crucial for effective treatment planning.
• Existing diagnostic methods can be time-consuming and may benefit from automated, AI-driven approaches.

Purpose of the Study:

• To develop and evaluate an automated lung cancer malignancy detection framework named ExtRanFS.
• To leverage transfer learning and advanced machine learning techniques for classifying lung CT scans.
• To improve the accuracy and efficiency of distinguishing between benign, malignant, and normal lung conditions.

Main Methods:

• Utilized the IQ-OTH/NCCD dataset comprising CT scans from 110 patients (40 malignant, 15 benign, 55 healthy).
• Employed a pre-trained VGG16 model as a convolutional feature extractor.
• Integrated an Extremely Randomized Tree Classifier for feature selection, followed by a Multi-Layer Perceptron (MLP) Classifier for final classification.

Main Results:

• The ExtRanFS framework achieved high performance metrics: 99.09% accuracy, 98.33% sensitivity, and 98.33% F1-Score.
• Comparative analysis demonstrated that the proposed framework outperformed other pre-trained models and existing state-of-the-art classifiers.
• The system effectively classified lung CT scans into benign, malignant, or normal categories.

Conclusions:

• The developed ExtRanFS framework shows significant promise for automated lung cancer malignancy detection.
• This AI-driven approach offers a highly accurate and efficient tool for medical practitioners.
• The system can aid in early and precise diagnosis, potentially benefiting patient treatment and outcomes.