A Novel Approach to Predict the Lungs Cancer Using the Hybrid Deep Learning ModelOpen Access

Abstract

Lung cancer is a predominant source of cancer-related mortality globally, and enhancing survival rates necessitates early detection. This research offers an integrated model that combines thresholding-based segmentation with convolutional neural networks (CNN) to predict lung cancer from CT scan pictures accurately. The initial step to isolate lung nodules from the background entails preprocessing the pictures by thresholding to generate binary images. During the second phase, a CNN is employed to analyze the segmented nodules and categorize them as benign or malignant. Our suggested model demonstrates substantial enhancements in critical performance parameters, including sensitivity (96%), recall (93%), precision (95%), and accuracy (97%), when compared to other models such as FCN, U-Net, and U-Net++. The results indicate the model's proficiency in reliably identifying and classifying lung nodules, positioning it as a potentially valuable instrument for early lung cancer diagnosis in clinical environments.