Back to all papers

Improving lung cancer detection with enhanced convolutional sequential networks.

Authors

Haziq U,Uddin J,Rahman S,Yaseen M,Khan I,Khan J,Jung Y

Affiliations (5)

  • Department of Computer Science, Riphah International University, Lahore, 55150, Punjab, Pakistan.
  • Department of Computer Science, University of Buner, Swari, 17290, Khyber Pakhtunkhwa, Pakistan.
  • Department of Computer Science, University of Engineering and Technology, Mardan, 32200, Khyber Pakhtunkhwa, Pakistan.
  • School of Computing, Gachon University, Seongnam, 13120, Gyeonggi-do, Republic of Korea. [email protected].
  • School of Computing, Gachon University, Seongnam, 13120, Gyeonggi-do, Republic of Korea. [email protected].

Abstract

Lung cancer is the most common cause of cancer-related deaths worldwide, and early detection is extremely important for improving survival. According to the National Institute of Health Sciences, lung cancer has the highest rate of cancer mortality, according to the National Institute of Health Sciences. Medical professionals are usually based on clinical imaging methods such as MRI, X-ray, biopsy, ultrasound, and CT scans. However, these imaging techniques often face challenges including false positives, false negatives, and sensitivity. Deep learning approaches, particularly folding networks (CNNS), have arisen as they tackle these issues. However, traditional CNN models often suffer from high computing complexity, slow inference times and over adaptation in real-world clinical data. To overcome these limitations, we propose an optimized sequential folding network (SCNN) that maintains a high level of classification accuracy, simultaneously reducing processing time and computing load. The SCNN model consists of three folding layers, three maximum pooling layers, flat layers and dense layers, allowing for efficient and accurate classification. In the histological imaging dataset, three categories of lung cancer models are adenocarcinoma, benign and squamous cell carcinoma. Our SCNN achieves an average accuracy of 95.34%, an accuracy of 95.66%, a recall of 95.33%, and an F1 score of over 60 epochs within 1000 seconds. These results go beyond traditional CNN, R-CNN, and custom inception classifiers, indicating superior speed and robustness in histological image classification. Therefore, SCNN offers a practical and scalable solution to improve lung cancer awareness in clinical practice.

Topics

Lung NeoplasmsDeep LearningNeural Networks, ComputerEarly Detection of CancerJournal Article

Ready to Sharpen Your Edge?

Join hundreds of your peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.