Performance Analysis of Optimizers on Convolutional Neural Network for Lung Cancer Classification Using CT Scan Images

Abstract

As a potentially fatal condition, lung cancer requires immediate medical attention. In particular, with imaging methods such as Computed Tomography scan imaging, early diagnosis at the early stage has the potential to save lives. To improve the accuracy of lung cancer classification, this study compares and evaluates several optimization algorithms, such as SGD, Adam, RMSprop, and AdaDelta. The goal is to determine how different optimizers affect the performance of CNN. The dataset utilized in this study was sourced from Kaggle. It included 5,958 CT scan pictures processed using a CNN model and divided into four classes: Adenocarcinoma, Squamous Cell Carcinoma, Normal and Large Cell Carcinoma. According to the trial data, the Adam optimizer had the highest accuracy, 72.22%, outperforming other optimization methods. This study contributes to advancements in AI for medical imaging, demonstrating that optimizer selection plays a crucial role in CNN-based lung cancer classification. The findings provide valuable insights for developing AI-assisted diagnostic tools to improve early lung cancer detection.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint