Peripheral Pulmonary Lesions Classification Using Endobronchial Ultrasonography Images Based on Bagging Ensemble Learning and Down-Sampling Technique

Lung cancer is the second most common in world, with an average five-year survival rate of 15 percent. Approximately 238,340 people were diagnosed US 2023 based on estimation American Cancer Society, and 127,070 died from it. has always been a big problem for scientists. There never good solution. So, early detection particularly important. In recent years, endobronchial ultrasonography (EBUS) images have used more diagnosis lung because their advantages real-time performance, no radiation, superior performance. This research aims to develop computer-aided (CAD) system differentiate benign malignant peripheral pulmonary lesions (PPLs). The efficacy this framework was evaluated dataset comprising 69 cases carcinoma, encompassing 59 instances 10 cases. final experimental results accuracy, F1-Score, AUC, PPV, NPV, sensitivity, specificity 0.7, 0.63, 0.75, 0.84, 0.68, 0.56, 0.85, respectively. From experiment results, developed CAD potential ability diagnose PPLs by using EBUS Deep Learning.