A Survey on Optimized CT Scan Analysis for Lung Cancer Detection Using Convolutional Neural Networks

Abstract

Lung cancer is a major cause of death around the world, so finding it early can help people get better treatment and have a better chance of surviving. Computer-Aided Diagnosis (CAD) systems have emerged as valuable tools in medical imaging, assisting in the early identification of lung tumors using Computed Tomography (CT) scans. However, accurate detection remains a challenge due to the irregular shape, varying location, and low contrast of tumors in CT images. To know these challenges the study presents a novel tumor and nodule segmentation model based on a Convolutional Neural Network (CNN). The model integrates preprocessing techniques such as image filtering for enhancement and postprocessing methods utilizing morphological operations for precise segmentation. Additionally, an active contour algorithm is employed to refine tumor boundary detection. We tested the model with sensitivity analysis on a benchmark dataset.

Introduction

Lung cancer is one of the most prevalent and deadliest cancers globally, accounting for 27% of cancer-related deaths in the U.S. (~350 deaths daily). Early detection is critical:

• Survival Rate: 87% if caught early, only 19% in late stages.

• CT Scans are a key tool but face issues like low contrast, complex lung anatomy, and tumor similarity to other lung structures.

To improve detection, Computer-Aided Diagnosis (CAD) systems using machine learning (ML) and image processing are becoming essential. However, detecting small nodules, a major early indicator, is still difficult—hence, the need for advanced segmentation and classification techniques.

II. Literature Review: Existing Approaches

1. Classical ML with Image Processing (2024)

   • Used LIDC dataset; image preprocessing (resizing, filtering, etc.).

   • Techniques like thresholding and morphological ops for segmentation.

   • Limitations: No deep learning, single dataset, lacks generalizability, no cross-validation.

2. MTMR-Net with Margin Ranking Loss (2019)

   • Multi-task deep network for both classification and semantic attribute scoring (texture, margin).

   • Based on ResNet and Siamese networks.

   • Limitations: Used 2D slices (not 3D), fixed attributes, interpretability issues.

3. 3D Probabilistic System with CNNs (2020)

   • End-to-end system using 3D CNNs, V-Net segmentation, and MIL for patient-level diagnosis.

   • Uses Bayesian modeling for uncertainty.

   • Limitations: Dataset lacks large nodule annotations; model is dependent on CADe performance; limited real-world testing.

4. Multi-task Deep Learning (MTMR-Net – 2019)

   • Same as #2, with detailed explanation of architecture (cross-entropy + MSE + margin ranking loss).

   • Limitations: 2D only, excluded ambiguous cases, dataset limitations, high computational cost.

5. Split-Method Using DL + Traditional ML (2024)

   • LDCT scans → DL feature extraction → ML classification.

   • Limitations: Data heterogeneity, image artifacts, limited generalization, high compute demand.

6. Segmentation Using Histopathology (2024)

   • Used U-Net + CNN ensemble (ResNet-50, VGG-16, EfficientNet-B5).

   • High accuracy but limited to histopathology images, not CT scans.

   • No clinical validation, high compute load.

7. CNN + GoogleNet on CT Scans (2022)

   • Achieved 98% precision for lung cancer detection.

   • Limitations: Limited dataset transparency, unclear on multiclass performance, lacks reproducibility info.

8. CNN + Ebola Optimization Search Algorithm (2023)

   • CNN + EOSA for classifying normal, benign, malignant.

   • Accuracy: 93.21%

   • Limitations: EOSA is untested, high complexity, lacks external validation.

9. DeepLung – 3D CNN (2017)

   • First fully automated 3D CNN pipeline for detection + classification.

   • Strength: Matches radiologist-level performance.

   • Limitations: Early model, lacks modern training techniques, no real-world integration.

10. Meta-Review on DL for Lung Cancer (Date unspecified)

• Reviewed various DL models on clinical imaging data.

• Highlighted issues like dataset inconsistency, publication bias, lack of external validation, and interpretability problems.

III. Proposed Methodology

The proposed system uses Convolutional Neural Networks (CNNs) on CT scans to detect lung cancer nodules (benign or malignant).

A. Preprocessing

• Challenges: Low contrast, artifacts in LDCT images.

• Techniques:

  • Gaussian filter: Smoothens images.

  • Sobel filter: Highlights edges (helps locate tumor boundaries).

B. Segmentation

• CNN architecture with 14 convolution layers + 7 activations.

• Designed to process small image patches.

• Uses:

  • Learning rate adjustment (starts at 0.0001, increases to 0.3)

  • Regularization to prevent overfitting.

  • Dense-to-convolutional transformations for final segmentation.

C. Tumor Identification

• Uses morphological operations (erosion, dilation, opening, closing).

• Helps isolate tumor structures based on shape and size.

• Focused on mimicking realistic tumor boundaries.

Conclusion

This approach effectively addresses key challenges in automated lung tumor sensing, like low image contrast, irregular tumor shapes, and noise among the scan data. Through the integration of image enhancement, CNN-based segmentation, morphological postprocessing, and final classification, the system achieves high levels of accuracy and sensitivity. The inclusion of performance evaluation metrics and optimization strategies makes the model not only technically sound but also clinically relevant. However, despite its strengths, the methodology could benefit from future enhancements such as incorporating 3D volumetric data instead of 2D slices, performing cross-dataset validation to improve generalizability, and reducing computational complexity for real-time deployment. Overall, this pipeline presents a robust and scalable approach for supporting radiologists in early and accurate lung cancer diagnosis.

References

[1] X.Zhouetal.,“Intelligent small object detection based on Digital twinning for smart manufacturing in industrial CPS,” IEEE Trans. Ind. Inform., vol. 18, no. 2, pp. 1377–1386, Feb. 2022, doi: 10.1109/TII.2021.3061419. [2] T. Saba, “Automated lung nodule detection and classification based on multiple classifiers voting,” Micro Res. Technique, vol. 82, no. 9, pp. 1601–1609, 2019. [3] I. Abunadi, “Deep and hybrid learning of MRI diagnosis for early detection of the progression stages in Alzheimer’s disease,” Connection Sci., vol. 34, no. 1, pp. 2395–2430, 2022. [4] J. Liu et al., “Automated cardiac segmentation of cross-modal medical images using unsupervised multi-domain adaptation and spatial neural attention structure,” Med. Image Anal., vol. 72, 2021, Art. no. 102135. [5] I. Abunadi and E. M. Senan, “Multi-method diagnosis of blood micro scopic sample for early detection of acute lymphoblastic leukemia based on deep learning and hybrid techniques,” Sensors, vol. 22, no. 4, 2022, Art. no. 1629. [6] K. Yildirim, P. G. Bozdag, M. Talo, O. Yildirim, M. Karabatak, and U. R. Acharya, “Deep learning model for automated kidney stone de tection using coronal CT images,” Comput. Biol. Med., vol. 135, 2021, Art. no. 104569. [7] A. J. Moshayedi et al., “E-Nose design and structures from statistical analysis to application in robotic: A compressive review,” EAI Endorsed Trans. AI Robot., vol. 2, no. 1, pp. e1–e1, 2023. [8] X. Zhou, W. Liang, I. Kevin, K. Wang, and L. T. Yang, “Deep correlation mining based on hierarchical hybrid networks for heterogeneous Big Data recommendations,” IEEE Trans. Comput. Social Syst., vol. 8, no. 1, pp. 171–178, Feb. 2021. [9] X.Zhou,W.Liang,K.WangandS.Shimizu,“Multi-modality behavioral influence analysis for personalized recommendations in health social media environment,” IEEE Trans. Comput. Social Syst., vol. 6, no. 5, pp. 888–897, Oct. 2019, doi: 10.1109/TCSS.2019.2918285. [10] A. Rehman, A. Naz, M. Khan, T. Saba, and K. A. Al-Ghamdi, “Microscopic lung tumor segmentation using convolutional neural network and morphological operations ,”Cancer Medicine, vol. 13, no. 13, pp. 12907–12919. [11] A. Raftarai, R. R. Mahounaki, M. Harouni, M. Karimi, and S. K. Olghoran, “Predictive models of hospital readmission rate using the improved AdaBoost in COVID-19,” in Intelligent Computing Applications For COVID-19. Boca Raton, FL, USA: CRC Press, 2021, pp. 67–86. [12] X. Zhou, X. Xu, W. Liang, Z. Zeng, and Z. Yan, “Deep-learning-enhanced multitarget detection for end–edge–cloud surveillance in smart IoT,” IEEE Internet Things J., vol. 8, no. 16, pp. 12588–12596, Aug. 2021. [13] X. Zhou, Y. Li, and W. Liang, “CNN-RNN based intelligent recommendation for online medical pre-diagnosis support,” IEEE/ACM Trans. Comput. Biol. Bioinf., vol. 18, no. 3, pp. 912–921, May/Jun. 2021, doi: 10.1109/TCBB.2020.2994780. [14] M. Pradhan, “An early diagnosis of lung nodule using CT images based on hybrid machine learning techniques,” in Machine Learning and Artificial Intelligence in Healthcare Systems. Boca Raton, FL, USA: CRC Press, 2023, pp. 311-329. [15] S. Jain, P. Choudhari, and M. Gour, “Pulmonary lung nodule detection from computed tomography images using two-stage convolutional neural network,” Comput. J., vol. 66, no. 4, pp. 785–795, 2023. [16] M. A. Jaffar, A. Hussain, F. Jabeen, M. Nazir, and A. M. Mirza, “GA-SVM based lungs nodule detection and classification,” in Proc. Int. Conf. Signal Process. Image Process. Pattern Recognit., Jeju Island, Korea, 2009, pp. 133–140. [17] S. T. Namin, H. A. Moghaddam, R. Jafari, M. Esmaeil-Zadeh, and M. Gity, “Automated detection and classification of pulmonary nodules in 3D thoracic CT images,” in Proc. IEEE Int. Conf. Syst., Man Cybern., 2010, pp. 3774–3779. [18] M. Tan, R. Deklerck, B. Jansen, M. Bister, and J. Cornelis, “A novel computer-aided lung nodule detection system for CT images,” Med. Phys., vol. 38, no. 10, pp. 5630–5645, 2011. [19] S. Sivakumar and C. Chandrasekar, “Lung nodule detection using fuzzy clustering and support vector machines,” Int. J. Eng. Technol., vol. 5, no. 1, pp. 179–185, 2013. [20] M. Keshani, Z. Azimifar, F. Tajeripour, and R. Boostani, “Lung nodule segmentation and recognition using SVM classifier and active contour modeling: A complete intelligent system,” Comput. Biol. Med., vol. 43, no. 4, pp. 287–300, 2013. [21] S. Akram, M. Younus Javed, U. Qamar, A. Khanum, and A. Hassan, “Artificial neural network based classification of lungs nodule using hybrid features from computerized tomographic images,” Appl. Math. Inf. Sci., vol. 9, no. 1, pp. 183–195, 2015. [22] F. Mahmudi, M. Soleimani, and M. Naderi, “Some properties of the maximal graph of a commutative ring,” Southeast Asian Bull. Math., vol. 43, no. 4, pp. 525–536, 2019. [23] J. Amin et al., “Brain tumor classification: Feature fusion,” in Proc. IEEE Int. Conf. Comput. Inf. Sci., 2019, pp. 1–6. [24] M. Keming and D. Zhuofu, “Lung nodule image classification based on ensemble machine learning,” J. Med. Imag. Health Inform., vol. 6, no. 7, pp. 1679–1685, 2016. [25] H. Madero Orozco, O. O. Vergara Villegas, V. G. Cruz Sánchez, H. D. J. Ochoa Domínguez, and M. D. J. Nandayapa Alfaro, “Automated system for lung nodules classification based on wavelet feature descriptor and support vector machine,” Biomed. Eng. Online, vol. 14, 2015, Art. no. 9. [26] E. E. Nithila and S. Kumar, “Segmentation of lung nodule in CT data using active contour model and Fuzzy C-mean clustering,” Alexandria Eng. J., vol. 55, no. 3, pp. 2583–2588, 2016. [27] E. Vorontsov, A. Tang, C. Pal, and S. Kadoury, “Liver lesion segmentation informed by joint liver segmentation,” in Proc. IEEE 15th Int. Symp. Biomed. Imag., 2018, pp. 1332–1335. [28] K. Suttitanawat, A. Uppanun, S. Auephanwiriyakul, N. Theera-Umpon, and P. Wuttisarnwattana, “Lung nodule detection from chest X-ray images using interval type-2 fuzzy logic system,” in Proc. IEEE 8th Int. Conf. Control System Comput. Eng., 2018, pp. 223–226. [29] N. Nasrullah, J. Sang, M. S. Alam, M. Mateen, B. Cai, and H. Hu, “Automated lung nodule detection and classification using deep learning combined with multiple strategies,” Sensors, vol. 19, no. 17, 2019, Art. no. 3722. [30] A. Halder, D. Dey, and A. K. Sadhu, “Lung nodule detection from feature engineering to deep learning in thoracic CT images: A comprehensive review,” J. Digit. Imag., vol. 33, no. 3, pp. 655–677, 2020. [31] M. A. Heuvelmans et al., “Lung cancer prediction by deep learning to identify benign lung nodules,” Lung Cancer, vol. 154, pp. 1–4, 2021. [32] M. Kanipriya, C. Hemalatha, N. Sridevi, S. SriVidhya, and S. J. Shabu, “An improved capuchin search algorithm optimized hybrid CNN-LSTM architecture for malignant lung nodule detection,” Biomed. Signal Process. Control, vol. 78, 2022, Art. no. 103973. [33] L. Agilandeeswari, S. D. Sree, and A. Bansal, “Deep learning and machine learning-based lung nodule detection systems–an analysis,” in Proc. 14th Int. Conf. Soft Comput. Pattern Recognit., 2023, pp. 215–225. [34] D. Khemasuwan, J. S. Sorensen, and H. G. Colt, “Artificial intelligence in pulmonary medicine: Computer vision, predictive model and COVID-19,” Eur. Respir. Rev., vol. 29, 2020, Art. no. 157. [35] M. Soleimani, M. H. Naderi, and A. R. Ashrafi, “Tensor product of the power graphs of some finite rings,” Facta Universitatis, Ser.: Math. Inform., vol. 34, pp. 101–122, 2019.

Copyright

Copyright © 2025 Hemalatha K, Harshita Manjunath Naik , Jay , Jeevitha S M, Ganesha G M. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.