Improving lung cancer detection with enhanced convolutional sequential networks

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Lung cancer is a leading cause of cancer mortality worldwide, necessitating improved early detection methods.
• Current medical imaging techniques for lung cancer diagnosis face limitations such as false positives and negatives.
• Traditional deep learning models, like Convolutional Neural Networks (CNNs), exhibit high computational complexity and slow inference.

Purpose of the Study:

• To develop an optimized sequential folding network (SCNN) for accurate and efficient lung cancer classification.
• To address the limitations of traditional CNNs in terms of speed and computational load.
• To enhance the practical applicability of deep learning in clinical lung cancer diagnosis.

Main Methods:

• The proposed SCNN model incorporates three folding layers, three maximum pooling layers, flat layers, and dense layers.
• The model was trained and evaluated on a histological imaging dataset containing adenocarcinoma, benign, and squamous cell carcinoma.
• Performance was compared against traditional CNN, R-CNN, and custom inception classifiers.

Main Results:

• The SCNN model achieved an average accuracy of 95.34% and an F1 score within 60 epochs.
• Classification accuracy reached 95.66% with a recall of 95.33%.
• The SCNN demonstrated superior speed and robustness compared to traditional CNN-based methods, completing classification within 1000 seconds.

Conclusions:

• The SCNN presents a practical and scalable solution for improving lung cancer detection accuracy and efficiency.
• This optimized deep learning approach offers a significant advancement over existing methods for histological image classification.
• The SCNN's performance suggests its potential to enhance clinical practice and patient outcomes in lung cancer diagnosis.