Optimizing Hybrid CNN-SVM Architectures for Enhanced Detection of Subsolid Lung Nodules: A Comparative Study with Traditional Methods

When observed on an X-ray or computed tomography (CT) scan, a nodule is a "spot on the lung." Actually, around one out of every 500 chest X-rays reveals a nodule. This little round or oval solid overgrowth of tissue is surrounded by normal lung tissue. A lung nodule is an oval or round clump of cells with a diameter less than 30 mm (1.2 inches). In the United States, nodules are quite prevalent and are found in around 1.6 million persons annually (Trusted Source). Both a single lung nodule and many nodules are conceivable. Identifying lung nodules early and correctly is key to enhancing the outlook for lung cancer and increasing the chances of survival for patients. This research introduces a combined Convolutional Neural Network-Support Vector Machine (CNN-SVM) model for spotting and categorizing lung nodules from CT scan images. The model produced good results for various types of nodules, with precision, recall, and F1-scores regularly surpassing 94%, and total accuracy of up to 99% for benign solid nodules. The use of confusion matrices and detailed performance indicators corroborated the model's dependability, demonstrating its capacity to deliver accurate early diagnosis of lung cancer.