Development of a machine learning-based multimode diagnosis system for lung cancer

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Oncology

Background:

• Artificial intelligence (AI) is increasingly utilized for identifying physical disorders.
• Lung cancer diagnosis remains a critical area for technological advancement.

Purpose of the Study:

• To develop and evaluate a novel three-layer AI-driven diagnostic system for lung cancer.
• To compare the efficacy of different machine learning algorithms in each diagnostic layer.

Main Methods:

• A three-layer diagnostic system was developed using Decision Tree C5.0, Artificial Neural Network (ANN), and Support Vector Machine (SVM).
• The Area Under the Curve (AUC) was used to assess the performance of each model.
• Layer 1: ANN (AUC=0.736) outperformed C5.0 (AUC=0.676) and SVM (AUC=0.640) using epidemiological data and clinical symptoms.
• Layer 2: ANN (AUC=0.889) and SVM (AUC=0.825) showed similar performance, both superior to C5.0 (AUC=0.804) with added tumor biomarkers.
• Layer 3: C5.0 achieved the highest AUC (0.910) and sensitivity (94.12%) using CT nodule-based radiomic features.

Main Results:

• The ANN model demonstrated superior performance in the initial screening layer.
• Subsequent layers showed improved diagnostic power, with ANN and SVM performing comparably in the second layer.
• The final layer, utilizing C5.0 with radiomic features, achieved the highest diagnostic accuracy and sensitivity for confirming lung cancer.

Conclusions:

• A robust three-layer AI system for lung cancer diagnosis was successfully developed.
• The system integrates different machine learning models strategically across layers for optimized screening and confirmation.
• This AI approach offers a promising tool for enhancing early detection and diagnosis of lung cancer.