Related Experiment Video
Updated: Sep 6, 2025
13:19
Deep Neural Networks for Image-Based Dietary Assessment
Published on: March 13, 2021
9.3K
Performance Analysis of State-of-the-Art CNN Architectures for LUNA16
Iftikhar Naseer1, Sheeraz Akram1, Tehreem Masood1
1Faculty of Computer Science & Information Technology, The Superior University, Lahore 54600, Pakistan.
Sensors (Basel, Switzerland)
|June 24, 2022
Summary
Convolutional neural networks (CNNs) are vital for complex ML tasks. AlexNet with SGD optimizer demonstrated superior performance in detecting lung cancer from CT scans, achieving 97.42% accuracy.
Area of Science:
- Artificial Intelligence
- Machine Learning
- Medical Imaging Analysis
Background:
- Convolutional Neural Networks (CNNs) are advanced machine learning tools for image recognition, NLP, and video analysis.
- CNN architectures are gaining significant attention for their potential in medical applications.
- Lung cancer detection remains a critical area in medical diagnostics.
Purpose of the Study:
- To compare the performance of various CNN architectures (LeNet, AlexNet, VGG16, ResNet-50, Inception-V1) for lung cancer detection.
- To evaluate the impact of different optimizers (RMSProp, Adam, SGD) on CNN performance.
- To identify the optimal CNN architecture and optimizer combination for accurate lung cancer diagnosis using CT images.
Main Methods:
- Utilized the publicly available LUNA16 dataset for training and validation.
- Implemented and compared five distinct CNN architectures: LeNet, AlexNet, VGG16, ResNet-50, and Inception-V1.
- Applied three optimizers: Root Mean Square Propagation (RMSProp), Adaptive Moment Estimation (Adam), and Stochastic Gradient Descent (SGD).
- Assessed performance using metrics including accuracy, specificity, sensitivity, positive predictive value, false omission rate, negative predictive value, and F1 score.
Main Results:
- The AlexNet architecture combined with the SGD optimizer achieved the highest validation accuracy of 97.42% for CT lung cancer detection.
- This combination also yielded a misclassification rate of 2.58%, sensitivity of 97.58%, and specificity of 97.25%.
- AlexNet with SGD outperformed other state-of-the-art CNN architectures in the comparative analysis.
Conclusions:
- AlexNet, when optimized with SGD, is highly effective for automated lung cancer detection in CT scans.
- The study highlights the importance of selecting appropriate CNN architectures and optimizers for medical imaging tasks.
- This research provides valuable insights for developing advanced AI-driven diagnostic tools for early lung cancer detection.
